Cohort study of effect of being overweight and change in weight on risk of coronary heart disease in old ageBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7097.1791 (Published 21 June 1997) Cite this as: BMJ 1997;314:1791
- Tamara B Harris, chief, geriatric epidemiology office ()a,
- Lenore J Launer, senior research scientistb,
- Jennifer Madans, acting associate director, vital and health statistics systemsc,
- Jacob J Feldman, associate director, office of analysis and epidemiology and health promotionc
- a Epidemiology, Demography, and Biometry Program, National Institute on Aging, Bethesda, MD, USA
- b Department of Chronic Diseases and Environmental Epidemiology, National Institute of Public Health and Environmental Protection, Bilthoven, Netherlands
- c National Center for Health Statistics, Hyattsville, MD, USA
- Correspondence to: Dr Harris
- Accepted 10 April 1997
Objective: To evaluate risk of late life coronary heart disease associated with being overweight in late middle or old age and to assess whether weight change modifies this risk.
Design: Longitudinal study of subjects in the epidemiological follow up study of the national health and nutrition examination survey I.
Setting: United States.
Subjects: 621 men and 960 women free of coronary heart disease in 1982-84 (mean age 77 years).
Main outcome measure: Incidence of coronary heart disease.
Results: Body mass index of 27 or more in late middle age was associated with increased risk of coronary heart disease in late life (relative risk=1.7 (95% confidence interval 1.3 to 2.1)) while body mass index of 27 or more in old age was not (1.1 (0.8 to 1.5)). This difference in risk was due largely to weight loss between middle and old age. Exclusion of those with weight loss of 10% or more increased risk associated with heavier weight in old age (1.4 (1.0 to 1.9)). Thinner older people who lost weight and heavier people who had gained weight showed increased risk of coronary heart disease compared with thinner people with stable weight.
Conclusions: Heavier weight in late middle age was a risk factor for coronary heart disease in late life. Heavier weight in old age was associated with an increased risk once those with substantial weight loss were excluded. The contribution of weight to risk of coronary heart disease in older people may be underestimated if weight history is neglected.
Little is known about the effects of being overweight (defined as a body mass index of ≥27) in old age on risk of heart disease
In this study older people who were overweight had an increased risk of coronary heart disease once weight history was accounted for
Weight history, particularly in late middle age, is important in assessing risk of coronary disease in older people
Older heavier people who gained more than 10% of midlife body weight or thinner older people who had lost 10% or more of body weight show high risk compared with thinner people with stable weight
Being overweight is an important contributor to risk of morbidity in younger people, particularly coronary heart disease.1 2 In people around age 65 heavier weight is associated with a modest increase in the risk of coronary heart disease,3 4 but it is unclear whether it is a health hazard for even older people.5 Clarification of risk associated with being overweight in late life is important as incidence of coronary artery disease increases with age, although it has been proposed that being overweight and weight gain with age carries little risk.5 In studies of weight change in old age, weight loss rather than weight gain was associated with coronary heart disease4 6 6 7 8 and weight change seemed to obscure the risk associated with being overweight.4 9
We examined the effects of being overweight and weight change on coronary heart disease in older men and women. Weights obtained roughly 10 years apart were used to estimate the risk of coronary heart disease associated with past and current weight, to determine the prevalence of weight change, to estimate the risk of coronary heart disease associated with weight change, and to evaluate whether weight change modified the estimation of risk associated with weight.
Subjects and methods
The epidemiological follow up study of the national health and nutrition examination survey I (NHANES I) is a longitudinal study of 14 407 people aged 25-74 in 1971-75.10 11 Follow up to 1987 included an in person interview in 1982-84 with measurement of weight, pulse, and blood pressure; a telephone interview in 1986 of people aged 55 or older at the time of participation in the original survey; and a telephone interview of all cohort members in 1987. Loss to follow up was less than 1% for those eligible.
For this paper we identified a cohort within the longitudinal study and followed them from 1982 to 1984. This design allowed us to use mainly measured weights to estimate change and more stringent criteria for excluding people with existing coronary heart disease. Mean length of follow up for outcome events was 3.9 (SD 0.8) years.
Our study population was aged 70 to 86 in 1982-84. We studied only white people since there were too few black participants to evaluate an interaction between race and weight. All subjects who had died before 1982-84 (n=940) or could not be located (n=3) were excluded. We also excluded 1166 people with coronary heart disease in 1982-84. This was defined as a report of heart failure, heart attack, or treatment for heart disease on interview in 1971-75; reporting at 1982-84 interview that a doctor had diagnosed a heart attack or angina; or discharge from hospital with a diagnosis of ICD-9 codes 410-414.
We excluded people with no data on weight or height in the original study and those with no weight recorded during 1982-84 (n=18). We did include reported weight for the 7% (110) who did not have weight measured in 1982-84. Mean body mass index was similar for those with estimated weight or with measured weight; reported weight has been shown to be highly correlated with measured weight.12 13 The final study population consisted of 1581 people: 621 men and 960 women.
New coronary heart disease (ICD-9 codes 410-414) in the cohort up to 1987 was determined from documentation of events in hospital discharge records (218, 83%) or from underlying cause of death on death certificates (45, 17%). There were 263 incident cases of coronary heart disease: 141 in women and 122 in men.
Procedures for measuring height and weight have been described.14 15 The 1982-84 measurements were considered as current and the 1971-75 measurements as past. We used the height measured in 1971-75 to calculate body mass index (kg/m2) for each point in time. These measures were pooled and sex specific cut off points based on thirds were calculated to be applied to both current and past body mass index, with those in the lowest third used as the reference group.16 An assessment of the effect of survival on these analyses showed that people excluded because of death were heavier in 1971-75 than those included in the cohort. Thus, our analysis is conservative since high risk heavier people have been excluded.
Weight change was calculated as percentage change from 1971-75 to 1982-84. This allows comparison of equivalent change among people of varying initial weight. We created three categories of weight change: gain of 10% or more, gain of less than 10% to loss of less than 10%, and loss of 10% or more. Those whose weight was relatively stable (gain or loss of less than 10%) were used as the reference group.
We tested the interaction between current body mass index and weight change and past body mass index and weight change. To examine how weight change affected risk of coronary disease associated with current weight, categories of weight change were crossed with categories of current body mass index to create nine analytical strata. We estimated risk of coronary heart disease in these nine strata using the thinnest weight stable group as the reference.
The risk factors for coronary heart disease included as covariates were systolic blood pressure, total cholesterol concentration (measured in 1971-75), and report of diabetes in 1982-84. Correlates of weight change included self rated health (excellent, very good, or good versus fair or poor health); alcohol intake (no intake versus any current alcohol intake); educational attainment (<9 years of education versus≥9 years); three variables estimating level of physical capacity (any difficulty versus no difficulty walking 400 m, level of usual or recreational physical activity (high, moderate, or low), or report of tiredness on usual activities); and report of serious illnesses including history of diabetes, stroke, or cancer.
The incidence of and mortality from coronary heart disease were calculated per 1000 person years of follow up. Results were similar for men and women so these groups were combined to increase the statistical power for subsequent analyses. Statistical modelling was performed with Cox's proportional hazards models to account for variable follow up time.17 All models were adjusted for sex, age, and cigarette smoking,16 although risks based on unadjusted crude rates were quite similar. Models of risk for weight change were adjusted for past body mass index.18 Final models were adjusted for available risk factors for coronary heart disease. The strength of associations was shown by transformation of ß coefficients to estimates of relative risk and calculation of 95% confidence intervals. Relation of variables with weight change were tested by χ2 association.
Since the study was restricted to a subgroup of the original cohort selected on characteristics which were included in sampling–that is, age and race19 20–we have presented unweighted results. Previous analyses of a group of similar age and race showed that weighting had little effect.
The 621 men and 960 women included had a mean age of 77 (SD 4) years in 1982-84. Cut off points for body mass index were 23.47 and 27.28 for women and 23.97 and 26.96 for men. More women (217, 23%) than men (76, 12%) lost 10% or more of body weight between the first and second measures while about equal proportions gained 10% or more (69 women (7%), 48 men (8%)).
Heavier body mass index in 1971-75 was associated with a 70% increase in the risk of coronary heart disease for both men and women (table 1). Heavier current body mass index was not associated with an increased risk of coronary heart disease in either men or women. Weight loss was associated with an increased risk of coronary heart disease for women (1.7 (95% confidence interval 1.2 to 2.4)) and for men (1.9 (1.2 to 3.1)). Weight gain was not associated with increased risk. Elimination of those who lost 10% or more of body weight increased the relative risk estimate among the heaviest third of current weight from 1.1 to 1.4 (1.0 to 1.9).
For men and women combined the risk associated with current weight was modified by history of weight change (multiplicative interaction term P=0.03). Risk of coronary heart disease was increased among heavier people who had gained weight and among thinner people who had lost weight (table 2).
We examined level of weight change by selected health characteristics in 1982-84 (table 3). Compared with weight stability, weight loss≥10% was related to history of diabetes and stroke, fair or poor health, and more complaints of tiredness. Those losing weight were also less physically active and reported more difficulty walking 400 metres. Those gaining weight had less education, were more likely to be former smokers, and reported more tiredness and difficulty walking distances. Controlling for these factors did not change the relation of weight and weight change with coronary risk.
Is heavier weight a risk factor for coronary heart disease in older people? We found a risk associated with being overweight, which suggests that coronary heart disease should be added to the several health risks already associated with heavier weight in very old age.21 22 We also found that knowledge of current weight may not be sufficient to identify those at risk in old age. Weight change from middle age to old age is also an indicator of risk, particularly for thinner people who had lost weight or heavier people who had gained weight. For survivors to old age, the relative importance of weight in middle age or weight in old age is unclear. However, several studies have shown weight in middle age (which may reflect maximum weight for most people) to be most important in determining risk of coronary heart disease. Without a long term weight history clinical assessment may underestimate the risk of disease in old age.4 9
We found that weight gain, while not associated with risk overall, was associated with increased risk of coronary heart disease in heavier people. An association of weight gain with increased risk of coronary heart disease is consistent with theoretical models of the effects of weight on risk factors for coronary heart disease risk23 and recent findings in younger women.24 Why weight loss is associated with an increased risk of coronary heart disease is unclear, since weight loss should decrease the level of cardiovascular risk factors.25 26 Although direct information on reasons for weight loss were not included in our data, sustained weight loss in old age is unlikely to be voluntary27 and may reflect other diseases, some of which might increase the risk of thromboembolic events. For instance, people with diabetes and stroke were disproportionately represented in the group that lost weight. Those with weight loss were also less likely to be physically active. Although controlling for these measures did not change the results, further investigation with more frequent weight measurements, reasons for weight loss, and physiological indicators of health would be useful.
Our case ascertainment was based primarily on diagnoses at hospital discharge.4 Reliance on the death certificate alone might be more likely to falsely attribute cases to coronary heart disease. However, exclusion of cases ascertained by death certificate only did not change our results. Silent myocardial infarction is common among older people and could result in missed cases. However, the prognosis for myocardial infarction is similar regardless of presentation.28 These cases of coronary heart disease should have been identified when patients were admitted to hospital for later cardiac complications or at the time of death.
There was also a potential for misclassification of current body mass index since current height is likely to be shorter than past height. Since height loss depends on unknown risk factors adjustment for mean height loss might introduce bias. We therefore elected to use the earlier measured value.
Heavier weight was a risk factor for coronary heart disease in this group of old men and women, although the association of current weight with risk was modified by weight history. Weight change was common, particularly weight loss. Weight history from middle age added to understanding health risk in late life. Neglect of weight history may lead to underestimation of the importance of being overweight as a risk factor in old age.
Funding: Data were collected under Federal inter-agency agreement.
Conflict of interest: None.