Article Text

Download PDFPDF

Co-occurrence of risk factors for cardiovascular disease by social class: 1958 British birth cohort
  1. C Power1,
  2. K Atherton1,
  3. O Manor2
  1. 1
    Centre for Paediatric Epidemiology and Biostatistics, UCL Institute of Child Health, London, UK
  2. 2
    School of Public Health and Community Medicine, Hebrew University, Jerusalem, Israel
  1. Professor C Power, Centre for Paediatric Epidemiology and Biostatistics, UCL Institute of Child Health, 30 Guilford Street, London WC1N 1EH, UK; c.power{at}ich.ucl.ac.uk

Abstract

Aim: To establish whether social differences in multiple risk factors for cardiovascular disease are due to a greater strength of association (higher correlation) between risk factors in less advantaged groups.

Methods: Co-occurrence of five risk factors (smoking, hypertension, low high-density lipoprotein cholesterol, obesity, diabetes) in 3614 British 45-year-old men and 3560 women in the manual and non-manual social groups.

Results: 4.0% of women in manual groups had ⩾3 risk factors compared with 1.7% in non-manual groups: 6.2% and 3.4% respectively for men. There was a higher than expected percentage of the population, overall, with ⩾3 risk factors assuming independence between risk factors; correspondingly, there was a slightly lower than expected proportion with one factor. However, patterns of observed to expected ratios were consistent in manual and non-manual groups and did not differ by the number of risk factors.

Conclusions: Higher prevalence of multiple risk factors in manual groups was due to the higher prevalence of individual factors rather than a greater tendency of those with an individual risk factor to have additional risks. Strategies to reduce multiple risk factors in less advantaged groups would help to lessen their health burden.

Statistics from Altmetric.com

It is known that risk factors for cardiovascular disease are correlated,14 and that multiple risk factors increase the risk of cardiovascular disease.510 Hence, disease risk scores, such as the Framingham score, have been developed to identify individuals most likely to develop cardiovascular disease.1113

The prevalence of cardiovascular disease risk varies by social position, with higher levels of adverse risk in less advantaged social groups,9 1417 and, furthermore, the prevalence of multiple risk is greater in less advantaged groups.3 10 1719 Although it is well-established that risk factors for cardiovascular disease tend to cluster, it is not yet clear whether the strength of this clustering differs by social group. Consequently, there is only limited understanding of whether the social difference in multiple risk is due to the higher prevalence of single risk factors or whether those in less advantaged social groups with one risk factor have a greater tendency than those in advantaged groups to have an additional risk factor.

Only a few studies have attempted to distinguish between these alternative explanations for the observed higher prevalence of multiple risk factors in less advantaged social groups and the evidence to date is inconsistent. A large study of 20–59-year-old men and women found that although there was a higher prevalence of multiple risks in lower educational groups, the association between risk factors did not differ by education level.19 Similarly, in a study of older women, aged 60–79 years, the association between risk factors in the Framingham score did not differ by socioeconomic position, in childhood or adulthood.10 However, in a younger population of 14 year olds there was some suggestion of differences between social groups, as indicated by income and cognition, in the strength of association between risk factors for cardiovascular disease.20

We aim to examine social differences in multiple risk factors for cardiovascular disease in mid-adulthood. Our objective is to establish whether social differences in multiple risk factors are due to stronger associations (eg, higher correlations) between risk factors in manual than non-manual social groups, that is, whether those in less advantaged groups with one risk factor have a greater tendency than those in advantaged groups to have additional risk factors. This greater tendency to have additional risk factors we refer to as “co-occurrence”, although others report on “clustering” of factors. We investigate multiple risk factors for cardiovascular disease in social groups, using information from a nationwide sample of British 45-year-old men and women.

METHODS

Study population

Participants were originally enrolled in the Perinatal Mortality Survey (PMS) of all born in England, Scotland and Wales, during 1 week in March 1958 with follow-up at intervals from birth to 45 years.2123 A total of 17 415 individuals participated in the PMS from an eligible sample of 17 638. Immigrants with the same birth dates were recruited to the study up to age 16 years (n = 920) to give a total of 18 558 cohort members. At age 45 years, excluding 1245 deaths and 1300 emigrants, there were 16 013 participants eligible for inclusion in the survey. However, the target sample invited to a clinical examination undertaken in the home by a trained nurse was 11 971 cohort members who had not died or emigrated, were still in contact with the survey, and who were deemed able to provide consent.24 A total of 9377 participants were examined, of whom 7174 had complete cardiovascular risk factor data (blood pressure, body mass index, high-density lipoprotein cholesterol (HDL-C), diabetes and smoking) and could be assigned a social class in adulthood.16 Ethical approval for the medical examination of the British 1958 Birth Cohort was obtained from South East Multi-Centre Research Ethics Committee; written consent was obtained from each participant at the 45 years clinical examination.

Measures

Risk factors in the calculation of the Framingham risk score were considered, namely blood pressure, body mass index, HDL-C, diabetes risk and smoking.11 25 Measurements of blood pressure, height, weight and non-fasting venous blood samples were collected at the 45-year biomedical survey. Three blood pressure measurements were taken with the participant seated for at least 5 minutes using an Omron 705CP automated digital oscillometric sphygmomanometer (Omron, Tokyo, Japan). A large cuff was used if the mid-upper arm circumference was ⩾32 cm. Measures deemed invalid by the nurse were excluded and an average of the remaining values was used in analyses. Hypertension was defined according to World Health Organization guidelines as systolic blood pressure ⩾140 mmHg, or diastolic blood pressure ⩾90 mmHg, or current antihypertensive medication.26 Height was measured using a Leicester portable stadiometer on a hard floor surface. Participants were measured unshod and stood upright with their head in the Frankfort plane. Weight was measured to the nearest 0.1 kg in light clothing with shoes removed. Where it was not possible to obtain an accurate measurement, or consent was not provided, self-reported weight (n = 100) or height (n = 84) was recorded. Obesity was defined as a body mass index (BMI) ⩾30 kg/m2, excluding pregnant women (n = 2).

HDL-C and triglyceride levels were measured using enzymatic methods by an autoanalyser (Olympus AU640, Japan); when the triglyceride level was >13 mmol/l, HDL-C levels were not measured. An HDL-C value of ⩽0.9 mmol/l was defined as low.27

Glycosylated haemoglobin (HbA1c) was measured on whole citrated blood by ion exchange high-performance liquid chromatography, using the Tosoh A1c2.2 Glycohemoglobin Analyser (HLC-723GHb; Tosoh, Tokyo, Japan). We defined individuals at risk of diabetes as having HbA1c levels ⩾7%, or those reporting a diagnosis of diabetes at 42 years, or those identified as taking diabetic medication at 45 years.

Smoking habit was measured at 42 years; individuals who reported that they currently smoked ⩾1 cigarette/day were defined as current smokers. Socioeconomic position was based on current or most recent occupation at 42 years (or 33 years if data were unavailable at 42 years; n = 737). Women were classified by their own occupation. Using the Registrar General’s social classification28, participants were grouped as non-manual (professional, managerial–technical and skilled non-manual classes) or manual (skilled, partly skilled and unskilled manual classes).

Analysis

The mean value (and SD) of continuous risk factors and the prevalence of each risk factor were calculated for non-manual and manual groups. Geometric means are presented for HDL-C and HbA1c. Correlation coefficients were calculated for continuous risk factors to compare associations in the manual and non-manual social groups. CIs for correlation coefficients were calculated using the Fisher Z-transformation. We assessed the co-occurrence of risk factors above that expected by chance as follows. The expected frequency of 0, 1, 2 or ⩾3 risk factors was calculated using the underlying prevalence of single risk factors and assuming that risk factors occurred independently of each other. To illustrate, an expected frequency of 0 risk factors (RFs) was calculated as the probability of not having any risk factor (NRF), that is NRF1×NRF2×NRF3×NRF4×NRF5 multiplied by the sample size n. The observed frequency of co-occurrence was divided by the expected frequency to give a ratio; observed to expected (O/E) ratios, which are somewhat high for no risk factors, low for single risk factors and high for ⩾3 risk factors indicate co-occurrence. Differences between observed and expected frequencies were assessed by χ2tests. Expected frequencies were calculated for non-manual and manual social classes and the O/E ratio for ⩾3 risk factors in social groups was compared using a z-type test. Because the prevalence and number with multiple risk factors varied by gender, all analyses were conducted separately for men and women. We repeated the analyses using social class at birth in addition to adult social class: the patterns of association were similar, hence the results are presented only for adult class.

Analyses were undertaken for 7174 participants with complete data on risk factors for cardiovascular disease and adult social class. We assessed the extent to which the sample with complete data resembled the original enrolment sample eligible for inclusion in this study (n = 16 013) in respect of social class at birth. In addition, we compared the complete data sample with all 45-year-old biomedical study participants in respect of other characteristics recorded prior to and at age 45 years. We next assessed the potential effect of any sample biases by applying weights in our main analyses for observed and expected risk factors. The sample of participants with complete data was weighted using appropriate weights to reflect the distribution of social class at birth (ie, for those eligible for inclusion at the 45-year survey). Then the analysis of observed and expected risk factors was repeated on the weighted sample, yielding similar results. Analyses were conducted using STATA 9.2 (StataCorp, Texas, USA) and Microsoft Office Excel 2003 (Microsoft, Washington).

RESULTS

Of 9377 participants at age 45 years, data were available for >96% for each risk factor for cardiovascular disease and social class separately; exceptions were HDL-C (83% with data) and diabetes (85% with data). Those with complete data for analysis in this study (n = 7174) did not differ from all 45-year biomedical survey participants (n = 9377) in respect of manual class at birth, respectively 70.6% versus 71.0%, whereas 73.5% of the eligible 45-year sample (n = 16 013) had a manual class of origin. (Further comparisons between the participating and eligible samples at 45 years are presented elsewhere.24) There were no differences between the sample with complete data for this study and all 45-year participants in respect of short stature at 7 years, own adult social class or prevalence of hypertension at 45 years. Negligible differences were found for own smoking (23.4% vs 24.2%, respectively), 45-year obesity (23.7% vs 24.6%) and diabetes (2.5% vs 3.0%).

Table 1 shows that the prevalence of smoking, hypertension, low HDL-C, obesity and elevated risk of diabetes was higher in manual social groups than in non-manual, in both sexes, although for men differences were not significant for HDL-C and elevated risk of diabetes. Social differences were also seen in the mean values for each separate risk factor.

Table 1 Mean (SD) or prevalence of risk factors for cardiovascular disease at 45 years in the 1958 birth cohort

In table 2 the strongest correlations observed for both sexes were between BMI and HDL-C, and between BMI and systolic blood pressure; for women but not men, correlations were of a similar magnitude between HbA1c and HDL-C and between BMI and HbA1c. However, the pattern of correlation was similar for manual and non-manual groups, and, allowing for multiple comparisons, coefficients did not differ significantly between social groups.

Table 2 Correlation coefficients (95% CIs)* for continuous risk factors. Values above the diagonal (non-shaded) are for non-manual class and below the diagonal (shaded) for manual class

Table 3 shows that 40% of all men and 51% of all women had none of the five risk factors examined. The distribution of the number of risk factors differed significantly between social groups, specifically 1.7% and 4.0% of women had three or more risk factors respectively in non-manual and manual groups; in men the percentages were 3.4% and 6.2%.

Table 3 Observed and expected prevalence of risk 0–3+ risk factors by adult social class, men and women separately

With respect to associations between risk factors, significant differences were found between observed and expected percentages by number of risk factors (table 3). The O/E ratio was somewhat greater than 1 for no risk factors, less than 1 for a single risk factor and much greater than 1 for ⩾3 risk factors, indicating that the proportion of individuals with ⩾3 risk factors was higher than expected assuming independence between risk factors, whereas the proportion with only one factor was slightly lower than expected. However, similar patterns were observed for manual and non-manual social groups; the excess of observed over expected for ⩾3 risk factors was not statistically significantly different between social groups (for men z = 1.41 p = 0.16; for women z = 0.88, p = 0.38). Results were largely unchanged in weighted analyses for O/E ratios and prevalence comparisons between non-manual and manual groups (data not presented).

To illustrate associations of risk factors by social groups, we examined co-occurrence of other factors among those with obesity and hypertension (437/3614 men and 219/3560 women). Among this group, the prevalence of low HDL-C was higher than expected, assuming independence in risk factors; however, the relative increase was similar for both manual and non-manual groups: in men with obesity and hypertension, 8.1% in manual groups had low HDL-C compared with 4.3% in all manual men (ratio of 1.9); among men in non-manual groups the prevalences are 7.0% and 3.8% (ratio of 1.8) respectively for those with obesity and hypertension compared with all non-manual men. Whereas, among those with obesity and hypertension, the prevalence of smoking was not higher than expected; this is seen in both manual and non-manual groups. For women in manual groups, 31.0% were smokers amongst those with obesity and hypertension compared to 32.3% expected; the corresponding figures for non-manual groups are 19.7% and 20.4%.

DISCUSSION

In this population sample of British adults in their mid-adulthood, a life stage at which clinical symptoms have yet to become common, 40% of men and 51% of women were identified as having none of five risk factors for cardiovascular disease. These risk factors were more prevalent among manual than non-manual social groups, such that multiple risk factors were more prevalent in the manual groups. But despite this difference in prevalence of multiple risk factors, our main finding is that the associations between a number of risk factors were similar for the social groups. Men had a higher prevalence of most risk factors, although the O/E ratio for 3 or more risk factors did not differ significantly by gender. Our finding of similar associations between risk factors in non-manual and manual social groups applies to both sexes.

Methodological considerations

A strength of this study is its large sample and nationwide coverage across Britain and, in addition, standardised protocols and equipment were used for measurement of risk factors for cardiovascular disease. The five risk factors included here (smoking, lipids, hypertension, diabetes and obesity) have been shown to account for a substantial percentage (about 80%) of the population attributable risk for acute myocardial infarction,7 highlighting the importance of these factors for cardiovascular disease outcome. Our risk factor measures correspond to those included in the Framingham studies in which coronary risk profiles were developed11 29 for use in a white population with no known coronary heart disease, hence our population fits these criteria. We lacked information on left ventricular hypertrophy, which although included in the risk score, has a low prevalence as seen for example in the Framingham and Renfrew and Paisley studies25 and in the British women’s heart and health study.10 To examine multiple risks we used cut-offs appropriate for our population, but which differ in some instances from those used elsewhere. For example, we used a systolic blood pressure threshold of 140 mmHg to define hypertension, lower than the threshold of 160 mmHg used in the British women’s heart and health study.10 Also, our definition of hypertension is based on casual blood pressure measures, which differs from a clinical definition, although such measurements in young and middle adulthood have been shown to predict subsequent cardiovascular mortality.30 31 For diabetes we identified those treated or reporting a doctor diagnosis grouped together with those at elevated risk on the basis of an HbA1c threshold of 7%,32 the latter being most often associated with a diagnosis of diabetes based on the oral glucose tolerance test.33 Obesity was assessed using the BMI, which may fail to distinguish between fat and lean mass, but is generally regarded as appropriate for capturing most of the relevant variation in overall levels of adiposity.34 35 Prevalence of some risk factors for cardiovascular disease was low and in order to have sufficient numbers for analyses we used a two-category grouping (manual and non-manual) for social class which allowed comparison of multiple risk factors. Risk profiles may not be fully developed by age 45 years, which may affect the generalisability of our findings. Finally, the loss to follow-up seen in the cohort over time could potentially affect the results of our study. Although differences between the analysis sample and those included in the 45-year biomedical survey were negligible, we found that more socially disadvantaged groups were less likely to participate at 45 years. Elsewhere we have identified the groups least likely to participate at 45 years, such as those with no male head of household at birth or those with experience of social care in early childhood.24 These patterns of response suggest that differences in the prevalence of risk factors between the manual and non-manual groups are potentially underestimated. And yet, comparison of weighted (to the social distribution at birth) estimates with prevalence levels reported for our analysis sample revealed only negligible differences; moreover, the ratios of O/E for manual and non-manual groups in the weighted analyses were little changed compared with those obtained from unweighted analyses.

Comparison of findings with other studies

The prevalence and correlation of risk factors for cardiovascular disease in our study population are broadly similar to other studies, although direct comparison is affected by differences in age of study samples, the risk factors considered and the cut-offs used to define high-risk. With respect to prevalence of risk factors, smoking was more common in our study than in some older populations,10 25 but comparison with 45–54-year-olds in the Health Survey for England in 2003 reveals many similarities.36 For example, 23% of men in the 1958 cohort were current smokers, identical to the prevalence for 45–54-year-old men in the Health Survey for England; 34% and 35% of men respectively were identified as hypertensive, and 25% and 28% respectively as obese. Major contributing factors to multiple risk in our study are smoking, obesity and hypertension, given that as expected for the age of our participants, only a minority have diabetes, in common with prevalence estimates from other UK studies,32 and the prevalence of low HLD-C is also low.

Our study is consistent with the literature demonstrating higher prevalence of individual and multiple risk factors for cardiovascular disease in less advantaged socioeconomic groups than in advantaged groups.3 10 1719 Interestingly, there is a tendency for stronger correlations between risk factors for cardiovascular disease in this middle-aged population than in older women10 and the correlation between BMI and systolic blood pressure is also stronger than in a US population in men, although not in women.1 Although our study is limited to a single, relatively young age (45 years), consideration of co-occurrence of risk factors at younger life stages is warranted given the observation that individual factors, such as smoking, adverse lipid profile and hypertension, have a greater relative effect on risk of acute myocardial infarction in younger than older individuals; furthermore, multiple risk factors account for a greater population attributable risk at younger ages consistently in men and women.7

With regard to associations between risk factors for cardiovascular disease in social groups, our finding that the higher prevalence of multiple risk factors in less advantaged social groups largely reflects the higher prevalence of individual risk factors rather than a greater tendency of those with one individual risk factor to have additional risks, is consistent with evidence from the few studies available to date. Thus, our conclusion based on a single age population, agrees with a Dutch study of 20–59-year-olds using a wider range of risk factors for cardiovascular disease and education level to indicate social position,19 and with a British study of 60–79-year-old women in which the association between risk factors for cardiovascular disease did not differ by socioeconomic position, in childhood or adulthood.10 Our results in the 1958 cohort were also seen with both social class in childhood and in adulthood. Social differences in clustering of related lifestyle factors (alcohol consumption, smoking, physical activity and diet) have been reported elsewhere,37 which may underlie, at least in part, the social differences in multiple risk factors for cardiovascular disease observed in our study, given that multiple lifestyle factors were found to be associated with an adverse lipid profile and high blood pressure in young adults.38

We found that men were more likely than women to have two or more risk factors (23.4% vs 13.7% respectively), as seen in some,3 19 but not all,8 previous studies. In some instances, the differences may be due to the populations studied. For example, a clinic-based study of middle-aged subjects with coronary artery disease found that women were more likely than men to have multiple risk factors,39 which contrasts with our findings in a community-based sample. Nonetheless, we found no consistent evidence that associations between risk factors varied by gender, in that the O/E ratios for number of risk factors did not differ for men and women. However, there was some evidence for stronger correlations between particular risk factors, notably for BMI and HDL-C and for BMI and systolic blood pressure, in both sexes; whereas in women associations of a similar magnitude were observed for BMI and HbA1c, and for HbA1c and HDL-C. It is recognised that absolute coronary risk is lower for women than men, but the relative risk associated with some factors may differ for men and women. In a large worldwide study, similar ORs were recorded in women and men for the association of acute myocardial infarction with smoking, raised lipids, abdominal obesity, but the increased risk associated with hypertension and diabetes was greater in women than in men.7 Synergistic effects of multiple risk factors may be greater in women, as suggested by findings from the Framingham offspring study,8 or from the San Antonio Heart study in which the risk of CVD death for women with diabetes and the metabolic syndrome was nearly three times higher than in men.40 However, in respect of social patterns of co-occurrence of risk factors, our findings apply similarly to both sexes.

Implication of findings

The higher prevalence of multiple risk factors for cardiovascular disease in less advantaged social groups in our study reflects the higher prevalence of individual risk factors rather than a tendency for those in manual groups to have greater co-occurrence of risk factors than those in non-manual groups. Given that the risk factors studied here are associated with an increased coronary heart disease burden, universally across population groups,7 our findings suggest that prevention of multiple risk factors in the least advantaged social groups is needed in order for social differences to be reduced. For this relatively young population, it remains to be seen whether mortality outcome relating to multiple risk factors varies by social position, as suggested by studies examining the prediction of risk scores in different social groups.25

What is already known on this subject

  • Socially disadvantaged groups have a more adverse cardiovascular disease risk profile than socially advantaged groups.

  • Risk factors for cardiovascular disease tend to co-occur or cluster. It is unclear whether social differences in multiple risk factors for cardiovascular disease are due to those in less advantaged groups with one risk factor have a greater tendency than those in advantaged groups to have additional factors.

What this study adds

  • The prevalence of multiple risk factors for cardiovascular disease was higher in manual than in non-manual social groups in mid-adulthood.

  • This social difference in multiple risk factors was due to a higher prevalence of individual factors in manual groups rather than a greater tendency of those in manual groups with an individual risk factor to have additional risks.

Policy implication

Policies to prevent multiple risk factors in less advantaged social groups are needed if social inequalities in cardiovascular disease risk are to be reduced.

Acknowledgments

We are grateful to the study participants in the 2002–2004 biomedical follow-up. The biomedical examination and related statistical analyses were funded by Medical Research Council awarded under the Health of the Public initiative. This work was undertaken at GOSH/UCL Institute of Child Health who received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme.

REFERENCES

Footnotes

  • Funding: Medical Research Council Health of the Public initiative (grant G000934).

  • Competing interests: None.

  • Ethics approval: Ethical approval for the medical examination of the British 1958 Birth Cohort was obtained from South East Multi-Centre Research Ethics Committee.

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.

Linked Articles

  • In this issue
    Mauricio L Barreto