METHODS

The subjects upon whom this research is based were offspring of 4064 married couples who were screened in the Renfrew-Paisley (Midspan) Study in 1972–76.^18,19 In 1993–4, attempts were made to trace offspring either through direct contact with the original study couple or, where both husband and wife were dead, through contacts with informants mentioned on their death certificate. Altogether 4829 offspring aged 30–59 were identified from 2365 couples with children; 3202 of these offspring lived within 45 minutes of Paisley and Renfrew. Of these, 1040 men and 1298 women from 1477 families completed a questionnaire and examination in 1996—a response rate of 73%. The study protocol was approved by the relevant research ethics committees.

Survey methods

Information on subjects' personal and family medical history, sociodemographic information, and smoking habits was obtained from detailed self completed questionnaires. Social class for women was based on their own, rather than their partner's, occupation. Examinations were carried out in clinics staffed by six research nurses who rotated between measurement stations. Measurements included: 12 lead ECG; blood pressure—recorded using a Dinamapp 8100, three readings were taken from the left arm with the subject seated and allowed to rest for five minutes—the mean of the last two reading for systolic and diastolic blood pressure was used; non-fasting venous blood samples were analysed for total and HDL cholesterol, fibrinogen,²⁰ white blood count and glucose; FEV₁ and FVC were measured using a Vitalograph Spirotrac III.²¹ Standing height and sitting height were measured without shoes in the Frankfort plane to the nearest millimetre using a Holtain stadiometer. Leg length was calculated as the difference between standing height and sitting height, plus the height of the stool. Weight was measured to the nearest 100 g with Seca digital scales in stockinged feet wearing indoor clothes. Recorded birthweight data were retrieved from birth archives for 677 (29%) study members.

Parents were examined between 1972–76 when height was measured to the nearest centimetre. We calculated mid-parental height as the average of mother's and father's height for each subject. Father's social class was determined from information given on regular occupation and classified according to the registrar general's classification. For retired men the last full time occupation was used. Occupational social class was classified as non-manual (social class I-III non-manual) and manual (III manual-V).

Statistical methods

All statistical analyses were performed using Stata version 6.0.²² Least squares regression was used to calculate age adjusted changes in continuous cardiovascular risk factor levels associated with a one standard deviation increase in subject's height, leg length, trunk length, leg:trunk ratio, and FEV₁. Likewise logistic regression was used to assess associations with dichotomous risk factors such as smoking and social class. All confidence intervals and p values were based on Huber White robust estimates of variance that accounted for intrafamilial clustering (P Huber, Proceedings of the Fifth Berkley Symposium on Mathematical Statistics and probability 1967).

We assessed the effects on FEV₁ risk factor associations of controlling for physical activity, measured by two questions in a self report questionnaire. The questions assessed levels of activity (a) during day to day activity and (b) during leisure time on 4 point scales.

RESULTS

Table 1 presents summary statistics for the cardiovascular risk factors examined. For most of the variables complete data were available for most subjects. The exceptions were recorded birth weight (available for 29% of subjects), HDL cholesterol (84%), and white cell count (89%). The social class distribution of survey members was slightly skewed towards the upper social classes because eligibility required that (a) the subject's parents had participated in 1972–76, (b) their parents were alive and traceable in 1993–4 (or traceable via the death certificate informant), and (c) the subjects themselves had agreed to take part in the study and non-response at each of the stages is likely to be socially patterned. Furthermore, in women, social class was based on their own occupations and many were in clerical employment.

Associations between anthropometry, FEV₁ and cardiovascular risk factors

Tables 2 and 3 shows the associations of height, leg length, trunk length, and FEV₁ with cardiovascular risk factors. To facilitate comparison across tables all coefficients represent the increase (or decrease) in the specified risk factor associated with a one standard deviation increase in each measurement. Greater overall height was associated with reduced total cholesterol and fibrinogen concentrations, non-manual social class (own and parental), higher FEV₁, and birth weight. In women, but not men, stature was associated with a lower white blood cell count and BMI.

Leg length, but not trunk length, was inversely associated with systolic and diastolic blood pressure. Leg length was more strongly inversely associated with cholesterol levels than trunk length. Greater trunk length was associated with a raised body mass index in men, the opposite was seen with respect to leg length in both sexes. High leg to trunk length ratios were associated with favourable profiles of blood pressure, BMI, HDL, and total cholesterol in all subjects as well as lower glucose in men and lower fibrinogen in women. Current smokers were more likely to have higher leg to trunk length ratios. Associations with other cardiovascular risk factors differed little for the two components of stature. Social class differences for leg length were similar to those for trunk length. In a separate analysis, based on the 677 subjects with recorded birthweight data, we repeated the analysis of leg length-risk factor associations, controlling for birth weight (data not shown). There was no evidence that birth weight confounded associations between leg length and cholesterol, fibrinogen or body mass index however the inverse associations of leg length with blood pressure was attenuated.

FEV₁ was more strongly associated with systolic blood pressure, fibrinogen, and white cell count than was height, leg length, trunk length or leg:trunk ratio. Unlike height, FEV₁ was associated with reduced random glucose and was more strongly associated with smoking. Associations with FEV₁ were essentially unchanged in models controlling for the possible confounding effects of physical activity (results not shown).

Effects of parental height on offspring's height-cardiovascular risk factor association

Controlling for mid-parental height had little effect on the associations between height and cardiovascular risk profile in their offspring (table 4). The association between stature and father's social class was, however, abolished in these models. This indicates that parental stature may act as a marker of parental socioeconomic position as well a marker of genetically determined height potential.

In separate models we assessed the effects of controlling for maternal and paternal height separately. Our findings were similar to those seen in models controlling for mid-parental height (see above, data not shown).

DISCUSSION

Main findings

This analysis suggests that taller men and women have more favourable cardiovascular risk profiles than shorter people. FEV₁ is more strongly associated than is height with the cardiovascular risk factors examined, suggesting it may be a better biomarker for the factors underlying associations between pre-adult exposures and adult cardiovascular disease.¹⁷ Associations with stature were largely unchanged in models controlling for parental height, indicating genetic factors may not underlie the observed height-risk factor associations. Furthermore, analysis of the subset of study members with birthweight data suggests that some stature-cardiovascular risk factor associations are independent of pre-natal influences.

Three important risk factors for CHD—raised blood pressure, BMI and, to a lesser extent, cholesterol—were more strongly associated with leg length than trunk length. Our findings for cholesterol and, to a lesser degree, blood pressure are consistent with those in our analysis of the Caerphilly cohort; in that analysis adverse patterns of insulin resistance and fibrinogen were also more strongly related to short leg length.¹⁵ In Midspan study members blood glucose levels were lower in men with higher leg: trunk ratios.

Unlike previous analyses,^15,23–25 we found no evidence that social class differences in height were in greater part attributable to leg length than trunk length. The associations of FEV₁ with leg length were slightly weaker than their relation with trunk length but were stronger than those we found reported in Caerphilly where we used a height adjusted measure of lung function.¹⁵ In the analyses reported here we used raw FEV₁ as we were interested in an absolute measure of lung development rather than lung function in relation to size.

Strengths and limitations

One of the main strengths of this study is the availability of detailed anthropometry (including sitting height) and lung function tests. Furthermore, the availability of parental height measurements enable an assessment of the importance of genetic influences in the height-CHD risk factor associations.

There are two main limitations to this analysis. Firstly, only 2338 (48%) of the offspring identified took part in the study. The main reason for non-participation was that these individuals had moved away from the Paisley and Renfrew area and so were not invited to take part (the response rate was 73% in those invited). We have no reason to believe that exposure-risk factor associations are likely to be different in the individuals examined compared with those who were not. In support of this suggestion, FEV₁-mortality associations are similar in the parents whose offspring participated in the study and those among parents whose offspring did not (Upton M, manuscript in preparation). Secondly, we had limited information on birth weight (29% of subjects only) and so were unable to fully assess its impact on observed associations in the full dataset. As previously reported,²⁶ associations of birth weight with leg length and trunk length were similar, suggesting birth weight is unlikely to confound the observed differential associations with leg length. However, there was some evidence that birth weight confounded the associations of leg length with blood pressure.

Early life and childhood influences on the development of CHD

The aetiology of CHD is complex and involves several pathological processes. The importance of development in utero and in infancy is indicated by associations of both low birth weight and weight at 1 year with a high risk of CHD.^27–29 The role of exposures in the first two decades of life in CHD development is supported by postmortem studies that reveal that around 50% of young men have some evidence of atherosclerosis (fatty streaks or plaques) in their coronary vessels.³⁰ Blood pressure, blood lipids, body mass index, and smoking in childhood predict the extent of such postmortem atherosclerotic changes.³¹ Just as birth weight is used as a biomarker for in utero exposures, our findings suggest that height and FEV₁ may be used as biomarkers for a range of exposures influencing post-natal growth such as diet, exposure to infection, and stress.^12,17 These factors in turn may have long term influences on the risk of developing atherosclerosis as indicated by height-CHD associations.^1–6,15 The distinct importance of childhood exposures, over and above those acting in utero, is supported by findings in two prospective studies where adjustment for birth weight did not greatly influence observed height-CHD associations.^2,3 While adult height is a marker for exposures influencing growth throughout childhood, leg length seems to be a marker for factors acting in the pre-pubertal period. Recent studies indicate that longer leg length is associated with breast feeding, energy rich diets at age 2 and affluent social circumstances.^25,24,32 Thus analyses reporting that leg length is the component of height most strongly related to later risk of CHD^7,15 indicate that one (or more) of these exposures pre-pubertally may be important in generating CHD risk.

FEV₁ and cardiovascular disease

Findings from the few studies^33–36 that have examined associations between lung function and cardiovascular risk factor levels are broadly consistent with those reported in this paper. In the Paisley and Renfrew cohort, which includes the parents of subjects forming the basis of our analyses, higher FEV₁ was associated with lower body mass index and cholesterol levels in men and women but there was no association with diastolic blood pressure.³³ Associations between FEV₁ and CHD mortality in the Paisley and Renfrew cohort were independent of other coronary risk factors including smoking, cholesterol, diastolic blood pressure, body mass index, and social class.³³ In Framingham, vital capacity was inversely associated with diastolic blood pressure, but associations with blood sugar and cholesterol differed in different age and sex groups.³⁵ In the Gothenburg cohort of middle age women, cholesterol and triglycerides levels decreased with increases in peak expiratory flow rates.³⁴

While previous research indicates that FEV₁ seems to be a more powerful predictor of CHD risk than height, height has two main advantages over FEV₁ as an epidemiological biomarker for early life exposures. Firstly, it is straightforward to measure accurately and is also self reported with reasonable precision. Secondly, with the exception of age related changes, it is little influenced by exposures subsequent to growth cessation. In contrast FEV₁ may be affected by smoking—itself a powerful predictor of CHD risk.

What this study contributes to an understanding of childhood influences on CHD

This analysis provides some support for the importance of childhood exposures in generating height-CHD associations. We have reported four main findings. Firstly, we have found associations between a range of well recognised cardiovascular risk factors and both height and FEV₁. Secondly, stature-risk factor associations are generally stronger for leg length than trunk length. This hints at the importance of pre-pubertal exposures, in generating height-CHD associations. Thirdly, controlling for parental stature had little effect on height-risk factor associations. This suggests that genetic factors may not underlie associations between height and cardiovascular risk factor profiles. Lastly, we found that FEV₁ was more strongly associated with cardiovascular risk factor profiles than the anthropometric measures. If FEV₁ is acting as a biomarker for childhood exposures in the FEV₁-cardiovascular risk profile associations, then this finding suggests it is a more sensitive measure of the childhood exposures, such as diet and socioeconomic circumstances, that lead to the development of CHD than is childhood stature.