Cardiometabolic risk is an umbrella term encompassing risk factors for CVD and type 2 diabetes, including increased TAG, fasting glucose, blood pressure, abdominal obesity and reduced HDL-cholesterol(Reference Brunzell, Davidson and Furberg1). An estimated seventy-six million US adults(Reference Ford2) and 2·9 million adolescents of 12–19 years of age(Reference Cook, Auinger and Li3) have the metabolic syndrome and the presence of any three of the above-mentioned five factors. The National Cholesterol Education Program Adult Treatment Panel III guidelines also include elevated LDL-cholesterol in its definition of the metabolic syndrome(4). A previous study has identified the metabolic syndrome as a significant predictor of the development of CVD and type 2 diabetes(Reference Gami, Witt and Howard5). Similarly, the Framingham Risk Score, which includes age, sex, cigarette smoking and the above-mentioned factors, was found to be associated with CHD and stroke(Reference Wannamethee, Shaper and Lennon6).
Metabolic, clinical and epidemiological studies have examined the influence of individual dietary and lifestyle factors on adverse cardiometabolic outcomes, showing that diets characterised by high polyunsaturated fat, low glycaemic index or glycaemic load, and increased consumption of fish, fruits, vegetables, nuts and whole grains appear to reduce the risk of cardiometabolic diseases(Reference Hu and Willett7). Furthermore, increased television (TV) viewing time(Reference Katzmarzyk, Church and Craig8–Reference Dunstan, Barr and Healy10), psychosocial stressors(Reference Rugulies11, Reference Wassertheil-Smoller, Shumaker and Ockene12), cigarette smoking(Reference Mitchell, Hwang and Vasan13, Reference Tamura, Tanaka and Okamura14), and both excessive drinking and abstinence from alcohol(Reference Kawano15) were associated with increased cardiovascular risk and mortality.
Healthy dietary and lifestyle patterns tend to co-exist and inter-relate with each other; therefore, it is important to examine dietary and lifestyle factors as a whole. However, only minimal data are available regarding the cumulative effects of healthy lifestyle behaviours on cardiometabolic risk. Stampfer et al. (Reference Stampfer, Hu and Manson16) examined the effect of a combination of modifiable factors including smoking, alcohol drinking, physical activity, BMI and diet on the risk of CHD among 84 129 women in the Nurses' Health Study. The outcome of interest was major coronary events which occurred rarely among subjects within the low-risk categories, making it hard to provide reliable estimates of the effects(Reference Stampfer, Hu and Manson16). A few studies have identified diet and lifestyle behaviours associated with the risk of heart failure(Reference Djoussé, Driver and Gaziano17) and hypertension(Reference Forman, Stampfer and Curhan18), but the effect of these behaviours on the clustering of cardiometabolic risk factors (rather than the actual disease) has not been evaluated.
The present study was based on a population-based, cross-sectional sample from the Lipid Research Clinic (LRC)'s Princeton Follow-up Study (PFS) to examine the effect of a combination of lifestyle behaviours (including fruit and vegetable intake, meat intake, smoking, alcohol consumption, physical activity and TV viewing time) on cardiometabolic risk reduction.
The analysis was based on 1454 participants from the PFS (1998–2003) who were followed up from the National Heart, Lung and Blood Institute LRC Prevalence Study in Cincinnati (1973–8). A detailed description of the LRC and PFS has been published elsewhere(Reference Tyroler19–Reference Friedman, Morrison and Daniels21). Briefly, the LRC was a multi-centre, multi-stage epidemiological study of lipids in eleven different populations in the USA and Canada. The LRC study in Cincinnati examined about 7000 students in grades 1–12 and 4500 parents in the Princeton School District. A family identification number was used to identify the familial correlation among parents and children from the same family. Stage 1 measured total cholesterol and TAG in all subjects plus demographic data. Stage 2 measured lipid profiles, blood pressure, BMI, triceps skinfolds, 24 h diet recalls, family history of CVD and twelve-channel clinical chemistries in random and hyperlipidaemic subsets of all stage 1 subjects. Stage 3 repeated stage 2 measures in all first-degree relatives of random and hyperlipidaemic index cases from stage 2, except for blood pressure.
After an average 26 years since their initial participation in the LRC study, eligible former schoolchildren were invited to participate in the PFS along with a sibling or a parent to assess the changes in lifestyles and cardiometabolic risk factors(Reference Tyroler19–Reference Friedman, Morrison and Daniels21). Among the 1454 PFS participants, 26 % were black and 56 % were female. There were 432 parents (age 66·5 (sd 6·6) years) and 1022 adult children (age 39·6 (sd 4·9) years).
The study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the Cincinnati Children's Hospital institutional review board. Written informed consent was obtained from all subjects. The University of Massachusetts Boston institutional review board approved the data analysis protocol.
Measures and definitions of healthy lifestyle and cardiometabolic risk in the Princeton Follow-up Study
Participants were assessed for sociodemographic characteristics, smoking, alcohol use, medical history and family history of certain diseases using standardised self-administered questionnaires. Dietary intake was measured by a Block-Quantified Food Frequency questionnaire. Information on physical activity was collected through a 7 d physical activity recall using the Paffenbarger Physical Activity Questionnaire, which had been validated against pedometer readings, treadmill tests and VO2max(Reference Paffenbarger, Wing and Hyde22–Reference Ainsworth, Leon and Richardson24). Participants' weight, height, waist circumference, blood pressure, LDL-cholesterol, HDL-cholesterol, total cholesterol, TAG, glucose and insulin were measured following standard procedures(Reference Laskarzewski, Morrison and Mellies25).
Diet measures chosen for inclusion in the present study were daily servings of meat, vegetables and fruits. Lifestyle factors were smoking (never/ever smoker), alcohol use (number of alcoholic beverages per week), physical activity (hours of moderate to vigorous physical activity (MVPA) per week) and TV viewing hours per d. Each of these six lifestyle factors has a substantial research base indicating potential reduction in cardiometabolic risk(Reference Wannamethee, Shaper and Lennon6–Reference Manson, Greenland and LaCroix9, Reference Wassertheil-Smoller, Shumaker and Ockene12–Reference Djoussé, Driver and Gaziano17).
Healthy lifestyle behaviour was defined based on public health recommendations and evidence from the previous literature on associations between individual lifestyle factors and cardiometabolic diseases. For fruit and vegetable intake, the Food Pyramid developed by the US Department of Agriculture recommends 3–5 servings of vegetables and 2–4 servings of fruits each day(26); therefore, a healthy behaviour for fruit and vegetable intake was defined as ≥ 5 servings/d. Because the Food Pyramid recommends 2–3 servings of meat daily(26), we defined a healthy behaviour for meat intake as ≤ 2 servings/d. The cut-off points for smoking status, alcohol consumption, TV viewing time and MVPA time were determined based on the findings from previous studies(Reference Manson, Greenland and LaCroix9, Reference Kawano15, Reference Djoussé, Driver and Gaziano17). For example, the healthy alcohol drinking behaviour was defined as at least 1–2 drinks/week. This is based on the finding that compared with non-drinkers ( < 1 drink/week), regular drinkers had a decreased risk of ischaemic stroke(Reference Kawano15). For TV viewing time, subjects with TV viewing time < 2 h/d had a lower CVD mortality compared with those with more than 2 h/d; therefore, ≤ 2 h TV viewing time was defined as a healthy behaviour(Reference Manson, Greenland and LaCroix9). For physical activity, previous studies found that at least 30 min of MVPA/d, corresponding to 3·5 h/week, was associated with a reduction in the risk of CHD and stroke(Reference Kawano15, Reference Djoussé, Driver and Gaziano17). The MVPA data in the present study were recorded in integers; therefore, we used 4 h/week as our cut-off point. We tested these cut-off points in exploratory data analysis, and found consistent relationships between each healthy lifestyle behaviour, as defined above, and cardiometabolic risk reduction. A combination of healthy lifestyle behaviours included daily fruit and vegetable intake ≥ 5 servings, daily meat intake ≤ 2 servings, never smoking, consuming 2–6 alcoholic drinks/week, TV viewing ≤ 2 h/d and MVPA ≥ 4 h/week. The presence of each behaviour was given a score of 1, so the healthy lifestyle score ranged from 0 to 6.
The primary outcome, cardiometabolic risk, was defined as having at least three of the following conditions: BMI ≥ 25 kg/m2, waist circumference ≥ 102 cm among men or ≥ 88 cm among women, blood pressure ≥ 130/85 mmHg or use of medication for hypertension; fasting blood glucose ≥ 1000 mg/l or use of glucose-lowering medication, total cholesterol ≥ 2000 mg/l, LDL-cholesterol ≥ 1300 mg/l, total TAG ≥ 1500 mg/l or drug treatment for elevated total TAG; and HDL-cholesterol < 400 mg/l in men or < 500 mg/l in women or drug treatment for reduced HDL-cholesterol. The cardiometabolic risk measures were based on recommendations from the Adult Treatment Panel III, the American Heart Association/National Heart, Lung and Blood Institute and the International Diabetes Federation(4, Reference Grundy, Cleeman and Daniels27, Reference Alberti, Zimmet and Shaw28). The secondary outcome of the present study was the composite cardiometabolic risk score, defined as an individual's total number of risk factors. This score varied between 0 and 9.
Means and standard deviations were calculated for continuous variables; numbers and proportions were calculated for categorical variables. A mixed-effect generalised linear regression model was used to evaluate associations of individual lifestyle factors with cardiometabolic risk in univariate analyses adjusted for within-family correlation identified by the same family identification number. Covariates including age, sex, race and generation were added to the model in multivariate analyses. OR and 95 % CI were obtained.
Next, each healthy lifestyle factor and the global healthy lifestyle score were examined as predictors of the presence/absence of cardiometabolic risk. The variables for individual healthy lifestyle factors were dichotomous (yes/no). The healthy lifestyle score was initially treated as a continuous variable, but only 1·2 % (eighteen) of the subjects had a healthy lifestyle score of 0, and only 2·6 % (thirty-eight) and 0·07 % (one) had scores of 5 or 6. Consequently, we collapsed the data into five groups to achieve robust sample sizes at the extremes: 0–1 as the reference group, 2, 3, 4 and 5–6. Four dummy variables were created for the five groups of the global healthy lifestyle score. An ordinal score was given to each category of the healthy lifestyle score, and then an analysis of P for a trend test was performed to examine the trend of cardiometabolic risk with increasing healthy lifestyle scores. Coefficients were obtained using the mixed-effect logistic regression because the outcome, cardiometabolic risk, was a binary variable.
We then used the composite score of cardiometabolic risk (i.e. the number of cardiometabolic risks) as the outcome variable. The composite score was categorised into four groups: 0–2 (reference group), 3–4, 5–6 and 7–9, so that each group had a similar sample size. Since the outcome was ordinal, the mixed-effect ordered logistic regression was used to model the likelihood of having 3–4 cardiometabolic risk factors, 5–6 cardiometabolic risk factors and 7–9 cardiometabolic risk factors, v. having 0–2 cardiometabolic risk factors (the reference group), on individual healthy lifestyle factors and then on the healthy lifestyle score, after adjustment for age, sex, race, generation and within-family correlation.
All P values were two-sided, and the significance level was set at 0·05. Data were analysed with Statistical Analysis Software version 9.1 (SAS Institute, Cary, NC, USA).
Table 1 summarises the prevalence of cardiometabolic risk according to demographic characteristics of the study population. Overall, 50·9 % (740 out of 1454) of the subjects had cardiometabolic risk. Males (53 %) and females (49·3 %) had a similar prevalence of cardiometabolic risk, as did Blacks (52·8 %) and Whites (50·2 %) (all P values>0·05). First-generation participants (parents) had a higher prevalence of cardiometabolic risk compared with their children, 66·9 % (289 out of 432) in parents v. 44·1 % (451 out of 1022) in their children (P < 0·001). Within generation, age was positively associated with cardiometabolic risk.
* Cardiometabolic risk was defined as having at least three of the following conditions: BMI ≥ 25 kg/m2, waist circumference ≥ 102 cm among men or ≥ 88 cm among women, blood pressure ≥ 130/85 mmHg or use of medication for hypertension; fasting blood glucose ≥ 1000 mg/l or use of glucose-lowering medication, total cholesterol ≥ 2000 mg/l, LDL-cholesterol ≥ 1300 mg/l, total TAG ≥ 1500 mg/l or drug treatment for elevated total TAG; and HDL-cholesterol < 400 mg/l in men or < 500 mg/l in women or drug treatment for reduced HDL-cholesterol.
The univariate and multivariate analyses of associations between individual lifestyle factors and cardiometabolic risk are summarised in Table 2. Daily fruit intake and frequency of alcohol consumption were associated with a decreased prevalence of cardiometabolic risk after adjusting for sex, age, race and generation, whereas daily meat intake, cigarette smoking and TV viewing time were associated with increased cardiometabolic risk (Table 2).
* P < 0·05
† Adjusted for intra-cluster correlation within family.
‡ Adjusted for age, race, sex, generation and intra-cluster correlation within family.
Then, each lifestyle factor was categorised dichotomously, generating six healthy lifestyle factors, including high fruit and vegetable consumption ( ≥ 5 servings/d), low meat consumption ( ≤ 2 servings/d), never smoking, consuming alcohol regularly (2–6 drinks/week), regularly participating in MVPA ( ≥ 4 h/week) and less TV viewing ( ≤ 2 h/d). About a third (437 out of 1454 or 30·1 %) of the participants had 2 healthy lifestyle behaviours, and 40·6 % (590 out of 1454) and 15·5 % (226 out of 1454) had 3 and 4 healthy lifestyle behaviours, respectively. The remaining 162 subjects (11·1 %) had 0 or 1 healthy lifestyle behaviour, and 39 (2·7 %) had 5 or 6 healthy lifestyle behaviours. As shown in Table 3, in the univariate model adjusted for the intra-cluster correlation within family, low meat consumption (OR 0·77; 95 % CI 0·61, 0·96), regular consumption of alcohol (OR 0·66; 95 % CI 0·52, 0·83), regularly participating in MVPA (OR 0·80; 95 % CI 0·63, 1·00) and less TV viewing time (OR 0·42; 95 % CI 0·23, 0·78) were inversely associated with cardiometabolic risk. These factors remained significant in multivariate analyses after adjustment for race, age, generation and sex, except for physical activity which became marginally significant. In addition, high fruit and vegetable consumption (OR 0·71; 95 % CI 0·55, 0·93) and never smoking (OR 0·69; 95 % CI 0·51, 0·92) were also inversely associated with cardiometabolic risk.
MVPA, moderate to vigorous physical activity; TV, television.
* P < 0·05
† Adjusted for intra-cluster correlation within family.
‡ Adjusted for age, race, sex, generation and intra-cluster correlation within family.
Next, we examined the healthy lifestyle score, which was the sum of the six healthy lifestyle factors. Fig. 1 shows that with each increment of the healthy lifestyle score, the prevalence of cardiometabolic risk decreased on average by 31 % (OR 0·69; 95 % CI 0·61, 0·78). A linear trend was observed. Compared with individuals having 0–1 healthy lifestyle score, those with 3 healthy lifestyle behaviours had a 47 % lower prevalence of cardiometabolic risk (OR 0·53; 95 % CI 0·36, 0·78), those with 4 healthy lifestyle behaviours had a 68 % lower cardiometabolic risk (OR 0·32; 95 % CI 0·20, 0·51), and those with 5 or 6 healthy lifestyle behaviours had a 70 % lower risk (OR 0·30; 95 % CI 0·13, 0·67). A higher healthy lifestyle score was associated with a lower prevalence of cardiometabolic risk (P for trend < 0·001).
Table 4 presents results from the ordered logistic regression on the association between healthy lifestyle factors and the composite score of cardiometabolic risk, which are further visualised in Fig. 2. There were statistically significant inverse relationships between the presence of healthy lifestyle behaviours and cardiometabolic risk score. For example, individuals who consumed alcohol regularly, compared with those who did not, had 27 % lower odds of a composite cardiometabolic risk score of 3–4 v. 0–2 (OR 0·73; 95 % CI 0·54, 0·98), and the protective effect of alcohol consumption increased as the cardiometabolic risk score increased. The findings were similar for the healthy lifestyle score. With every increment in healthy lifestyle score, the odds of a cardiometabolic risk score of 3–4 v. 0–2 were 18 % lower (OR 0·82; 95 % CI 0·72, 0·95), the odds of having the score as 5–6 v. 0–2 were 32 % lower (OR 0·68; 95 % CI 0·59, 0·79), and the odds of having the score as 7–9 v. 0–2 were 36 % lower (OR 0·64; 95 % CI 0·53, 0·78).
MVPA, moderate to vigorous physical activity; TV, television.
* P < 0·05.
† Adjusted for age, race, sex, generation and intra-cluster correlation within family.
Results from the present study indicated that a combination of healthy lifestyle factors including sufficient fruit and vegetable intake, less meat intake, abstinence from smoking, moderate drinking, regular exercise and less TV viewing time was associated with a decreased cardiometabolic risk. Each additional healthy lifestyle factor was associated with a 31 % reduction in cardiometabolic risk. Cardiometabolic risk decreased in a graded manner as the number of healthy lifestyle factors increased. These results support the hypothesis that reductions in cardiometabolic risk are associated with both individual and cumulative adherence to healthy lifestyle behaviours(Reference Lloyd-Jones, Hong and Labarthe29). When individual lifestyle factors were examined individually, most, but not all, factors were associated with cardiometabolic risk reduction.
The lifestyle factors in the present study have been examined individually or jointly by other researchers, and the present study adds to this literature by focusing on precursors to the actual disease rather than on frank disease outcomes as has typically been the case. For example, Chiuve et al. (Reference Chiuve, McCullough and Sacks30) found that a low-risk lifestyle including healthy diet, not smoking, alcohol consumption 5–30 g/d, MVPA ≥ 30 min/d and maintaining BMI < 25 kg/m2 resulted in a 62 % reduction in CHD among men of middle age or later in life. Djoussé et al. (Reference Djoussé, Driver and Gaziano17) found that a combination of BMI < 25 kg/m2, not smoking, exercise ≥ 5 times/week, alcohol intake ≥ 5 drinks/week, breakfast cereal ≥ 1 serving/week and fruits and vegetables ≥ 1 serving/week reduced the risk of heart failure. In another study, adherence to the Mediterranean diet, no smoking and exercise were associated with reduced total cholesterol, LDL-cholesterol and TAG(Reference Sofi, Gori and Marcucci31). In addition, several studies have documented the effects of other lifestyle factors such as folic acid supplementation(Reference Forman, Stampfer and Curhan18, Reference Mangoni, Sherwood and Swift32, Reference van Dijk, Rauwerda and Steyn33), dairy consumption(Reference Moore, Singer and Bradlee34), whole-grain intake(Reference Liu, Manson and Stampfer35) and psychosocial stressors(Reference Yusuf, Hawken and Ounpuu36). To the best of our knowledge, no studies have examined the effect of a combination of healthy lifestyle factors on cardiometabolic risk, which is a precursor of CVD and type 2 diabetes(Reference Wilson, Agostino and Parise37). By targeting lifestyle behaviours associated with cardiometabolic risk, we can prevent and manage risk for cardiometabolic disease and adverse outcomes of these disease processes at the early stages of development.
The present study found that fruit and vegetable intakes, when viewed on a continuous scale, were not significantly associated with cardiometabolic risk, whereas a high intake of fruits and vegetables ( ≥ 5 servings/d) was associated with reduced cardiometabolic risk. This suggests a possible threshold effect of fruit and vegetable consumption on cardiometabolic health. The present findings are consistent with the results of a previous epidemiological study, which found a similar protective effect of a high fruit and vegetable intake. For example, in a study in the USA, compared with people consuming < 3 servings of fruits and vegetables per d, those consuming 5–7 servings per d had a 73 % lower risk for myocardial infarction(Reference Rimm, Ascherio and Giovannucci38). A study in India found that people consuming 3·5 servings of vegetables per d had a threefold lower risk of IHD than did those consuming 0·8 servings/d(Reference Rastogi, Reddy and Vaz39).
Strengths of the present study include the population-based dataset, the objective measurement of cardiometabolic risk and the robustness of the results from different analyses. Healthy lifestyle factors were tested individually and jointly as a composite healthy lifestyle score, while the outcome, cardiometabolic risk, was measured both dichotomously and as a composite risk score. The results were largely consistent across the analyses, corroborating the evidence of the effect of a healthy lifestyle on cardiometabolic risk. The finding that MVPA was not individually associated with cardiometabolic risk reduction and showed no discernible dose–response pattern is surprising in the light of substantial literature to the contrary and may indicate a low validity of the MVPA data (see later for further discussion on a possible recall error in physical activity data). The absence of a statistically significant effect of cigarette smoking, conversely, probably reflects the sample size given the width of the CI.
The study is limited, however, by the cross-sectional design, which does not allow us to establish the time sequence of exposure and outcome, therefore preventing us from inferring causality. The selection bias in the cross-sectional design may also affect the estimated associations from the present study. Participants with a very unhealthy lifestyle and lethal cardiometabolic outcomes may not be included in the study, which is likely to lead to an underestimation of the effect of the lifestyle factors on cardiometabolic risk. A prospective, longitudinal study is desired to obtain more conclusive evidence on the temporal relationship between lifestyle and cardiometabolic risk.
Physical activity was measured by 7 d physical activity recall using the Paffenbarger Physical Activity Questionnaire in the present study(Reference Paffenbarger, Wing and Hyde22). Physical activity is a complex behaviour with four major components, frequency, intensity, duration and mode, which makes it difficult to obtain accurate measurement of all these components(Reference Heyward40). The 7 d recall method is practical and feasible for administering to large populations of individuals; however, due to its subjective nature in methodology, it is subject to recall error and unable to quantify physical activity with sufficient accuracy and precision. When compared with the doubly labelled water method, which is the criterion method to evaluate physical activity-related energy expenditure, this 7 d recall method has a low accuracy in the estimation of physical activity-related energy expenditure for individuals under free-living conditions. In addition, it is imprecise with an underestimation of physical activity-related energy expenditure in the most-active individuals and an overestimation in the least-active individuals(Reference Leenders, Sherman and Nagaraja41). It was found that MVPA was not individually associated with cardiometabolic risk reduction in the present study, indicating a low validity of the MVPA data measured by the 7 d recall method.
Another limitation of the present study is that some potential confounders such as education and income plus unknown confounders were not adjusted for, so a potential confounding effect could not be ruled out. Future study should use a longitudinal study design to examine the impacts of lifestyle in early life on cardiometabolic risk in later life. In addition, diet appears to be important for cardiometabolic risk in the present study, but we used only three aspects of dietary intake – fruits, vegetables and meat. Other dietary components such as total energy intake, fat, sugar, fibre intake, etc., which may influence CVD risk, were not included in the analyses.
Results from the present study provide further evidence supporting the beneficial effect of healthy lifestyle behaviours on cardiometabolic risk. It has important public health implications regarding the prevention of cardiometabolic disease.
L. S., J. A. M., J. W. and L. L. H. designed the study; L. S. performed the statistical analyses and wrote the manuscript; M. H. conducted the literature review; J. A. M., L. L. H. and J. W. edited the manuscript and helped with the interpretation and discussion of results. All authors approved the final manuscript. The authors do not have any conflicts of interest. The present study was supported by research grant HL62394 from the National Institutes of Health (J. A. M.) and the Joseph P. Healey from the University of Massachusetts Boston (L. S.).