CVD is a severe public health threat around the world, which mainly includes CHD, cerebrovascular disease, peripheral arterial disease and rheumatic heart disease. According to the WHO, an estimated 17·9 million people died from CVD in 2016, accounting for 31 % of all deaths globally. The Global Burden of Disease Study also announced that the total deaths from CVD increased by 21 % between 2007 and 2017(Reference Roth, Abate and Abate1).
Obesity is an independent risk factor for CVD and associated with many co-morbidities of CVD such as hypertension, hyperlipidaemia and the metabolic syndrome(Reference Ortega, Lavie and Blair2). As a serious public health issue worldwide, obesity had affected a total of 107·7 million children and 603·7 million adults from 195 countries in 2015, resulting in 4·0 million (2·7–5·3 million) deaths and 120 million (84–158 million) disability-adjusted life years among adults globally. CVD was the leading cause of deaths and disability-adjusted life years related to obesity(Reference Afshin, Forouzanfar and Reitsma3). It is generally known that BMI is a common indicator to define obesity, which is easy to perform. However, it does not reflect the location of fat distribution, and conclusions about the relationship between BMI and CVD are inconsistent. Some studies found a protective effect of BMI on CVD, which is called ‘obesity paradox’(Reference Romero-Corral, Montori and Somers4,Reference Lavie, Milani and Ventura5) . Coutinho et al. found that the risk of death decreased with the increase of BMI in 15 923 patients with coronary artery disease, whereas abdominal obesity was positively associated with the mortality(Reference Coutinho, Goel and Correa de Sa6).
Some indices of abdominal obesity such as waist circumference (WC), waist:hip ratio (WHR) and waist:height ratio (WHtR) are considered to be good indicators for predicting CVD. These indicators can reflect fat distribution and visceral fat accumulation to some extent, which are more easily available than computed tomography and MRI. Many epidemiological studies have begun to use these indicators to assess the relationship between obesity and CVD accurately. The international day for the evaluation of abdominal obesity, a large epidemiological study, investigated the relationship between WC and risk of CVD in 168 000 primary care patients in sixty-three countries, which demonstrated a significant correlation between waistline expanding and increased risk of CVD(Reference Balkau, Deanfield and Despres7). Karin et al. found that WHtR is a useful indicator to estimate the risk of CVD in patients with diabetes(Reference Radholm, Chalmers and Ohkuma8). The meta-analysis of Ashwell et al.(Reference Ashwell, Gunn and Gibson9), Lee et al. (Reference Lee, Huxley and Wildman10) and Savva et al.(Reference Savva, Lamnisos and Kafatos11) also suggested that measures of abdominal obesity were superior to BMI in detecting cardiovascular risk. However, studies from different countries and ethnicities have different conclusions regarding the superiority of different obesity indicators. Therefore, we collected the related prospective studies and made a meta-analysis to estimate the relationship between abdominal obesity and CVD by WC, WHR and WHtR.
Methods
Search strategy
The meta-analysis was conducted by the Meta-analysis of Observational Studies in Epidemiology. We conducted a systematic search of the literature on prospective cohort studies from PubMed, Embase and Web of Science up to 28 September 2019, using the following keywords and their synonyms (combined unexploded version of the Medical Subject Headings): (‘cardiovascular diseases’ OR ‘cerebrovascular disease’ OR ‘coronary heart disease’ OR ‘stroke’ OR ‘myocardial infarction’ OR ‘venous thromboembolism’) AND (‘waist circumference’ OR ‘waist to hip ratio’ OR ‘waist to height ratio’) AND (‘prospective cohort study’ OR ‘follow-up study’). The reference lists within the relevant publications were searched to identify other additional information. The corresponding author was contacted to request the required data if necessary.
Inclusion criteria
Studies were included in the meta-analysis if they met the following criteria: (1) participants were 18 years or older; (2) follow-up durations were more than 3 years; (3) at least one of the anthropometric measures was measured and reported, including WHR, WC or WHtR during the follow-up; (4) the outcome was the occurrence of CVD; (5) risk ratios (RR) or hazard ratios and their corresponding 95 % CI could be calculated and (6) studies were prospective study design.
Data abstraction and assessment for study quality
From each retrieved study, we extracted the following information: the first author’s name, year of publication, specific outcomes, name of cohort, number of participants/cases, mean age or age range, follow-up duration, anthropometric measurement method, covariates adjusted in multivariate analysis, quintiles of WHR, WC and WHtR, and corresponding RR of CVD with 95 % CI. Quality of the included studies was evaluated according to the Newcastle–Ottawa scale for non-randomised studies. The score of 0–3, 4–6 or 7–9 was regarded as low, moderate or high quality, respectively. Data extraction process and study quality assessment were performed by two independent investigators (X. R. and G. Y.), and group discussion to solve disagreements.
Statistical analysis
We collected the RR with 95 % CI as the common measure of associations across studies, where necessary, the hazard ratios were used to approximate RR. For the comparison between the highest categories and lowest categories of WC, WHR and WHtR, the summarised RR or hazard ratios and their corresponding 95 % CI were calculated using a random effects model. The model with the greatest control in each study was used to eliminate potential confounding factors.
Dose–response analysis was conducted using the method described by Greenland & Longnecker(Reference Phillips12) based on an increase of 10 cm for WC, 0·1 units for WHR and WHtR. We extracted the categories of WC, WHR and WHtR, the distributions of cases and number of subjects or person-years, and RR with 95 % CI. If studies reported results separately for different outcomes or other subgroups besides men and women, we combined the subgroup-specific estimates using a fixed effects model to generate an overall estimate so that each study was only conducted once in the main analysis. If the reference category was not the lowest in some studies, the RR were recalculated with the lowest category as a reference by the method provided by Hamling et al.(Reference Hamling, Lee and Weitkunat13). The median point in each category of WC, WHR and WHtR was assigned to the corresponding RR for each study. If medians were not reported, we considered the midpoint of the upper and lower boundaries as the dose of each category. If the highest and/or lowest category was open-ended, the midpoint of that category was set by assuming that the categorical width was the same as the next adjacent category.
Heterogeneity of studies was estimated by the I 2 statistic, and the values of 25, 50 and 75 % were defined as low, moderate and high heterogeneity, respectively. We performed subgroup analysis to evaluate the potential sources of heterogeneity. The analyses were conducted in men and women, respectively, to rule out the potential effects of sex difference. Subgroups analysis was carried out according to outcomes, age, geographic location, follow-up durations, number of participants and quality of articles. Potential publication bias was assessed using Begg’s test, Egger’s linear regression test and funnel plot asymmetry. If publication bias existed, ‘trim-and-fill’ was applied to correct the bias(Reference Duval and Tweedie14). We conducted sensitivity analysis in which a single study was excluded from the analysis at a time to verify the reliability of this study. All statistical analyses were performed using Stata statistical software version 12 (Stata Corp). P < 0·05 was considered as significant.
Results
Literature search and study characteristics
We found 6580 potential articles through literature search, and 6399 articles were excluded by title and abstract, leaving 181 relevant articles for review of full text. Thirty articles were left after full-text review according to the exclusion criteria (details are shown in Fig. 1). One of the articles included two different cohorts was regarded as two independent studies(Reference Gelber, Gaziano and Orav15). Finally, thirty-one prospective cohort studies and a cross-sectional study with follow-up were included in this meta-analysis with a total of 669 560 participants and 25 214 cases. There were five studies for female participants(Reference Gelber, Gaziano and Orav15–Reference Zhang, Shu and Gao19), eleven studies for male participants(Reference Gelber, Gaziano and Orav15, Reference Rimm, Stampfer and Giovannucci20–Reference Myint, Kwok and Luben29) and fifteen studies for both male and female participants(Reference Abete, Arriola and Etxezarreta30–Reference Larsson, Wolk and Hakansson44). Mean age ranged from 34 to 75 years old. The characteristics of included studies are presented in Table S1.
Fig. 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. WC, waist circumference; WHR, waist:hip ratio; WHtR, waist:height ratio; HR, hazard ratio; RR, risk ratio.
Waist circumference and CVD
Twenty-four studies examined associations between WC and risk of CVD, including 571 789 participants and 18 452 cases. Compared the highest category with the lowest category, the summary RR was 1·43 (95 % CI, 1·30, 1·56; I 2 = 58·6 %; P < 0·001). The pooled RR with 95 % CI was 1·42 (95 % CI 1·25, 1·61) and 1·46 (95 % CI 1·25, 1·72) in men and women with moderate heterogeneity, respectively (men: I 2 = 55·3 %, P heterogeneity < 0·05; women: I 2 = 61·4 %, P heterogeneity < 0·05) (Fig. 2). Eleven studies were available for inclusion in dose–response meta-analysis. Through sex-specific linear dose–response analysis, we found that each 10 cm increase of WC enhanced the risk of CVD by 4·0 % in men (RR 1·04; 95 % CI 1·02, 1·06; P < 0·001) and 3·4 % in women (RR 1·03; 95 % CI 1·01, 1·06; P < 0·05). The non-linear dose–response analysis also showed a significant relevance between WC and risk of CVD in both men and women (P < 0·001). In male participants, we observed that there was no change in the trend of the curve when WC was lower than 80 cm. When WC was about 80–90 cm, the incidence of CVD grew slowly. However, it was obvious that the risk of CVD increased with a steep slope when WC was over 90 cm. In women, the risk of CVD increased significantly when WC was over 80 cm (Fig. 3).
Fig. 2. Forest plot of association between waist circumference and risk of CVD. ES, effect size.
Fig. 3. Dose–response analysis between waist circumference (cm) and risk of CVD among men (left) and women (right). Solid line: non-linear dose response. Dotted line: 95 % CI. Straight dotted line: linear dose–response.
Waist:hip ratio and CVD
Twenty-one studies with 447 467 participants and 16 828 cases were included in the analysis of WHR and risk of CVD. The summary RR was 1·43 (95 % CI 1·33, 1·54; I 2 = 30·6 %; P < 0·001) compared the highest category with the lowest category. The pooled RR was 1·43 (95 % CI 1·31, 1·57) with a low heterogeneity (I 2 = 17·0 %; P heterogeneity = 0·259) in men, and 1·48 (95 % CI 1·29, 1·70) with a moderate heterogeneity (I 2 = 49·9 %; P heterogeneity = 0·02) in women (Fig. 4). Twelve studies provided enough information to carry on dose–response meta-analysis. The risk of CVD increased by 4·0 % in men (RR 1·04; 95 % CI 1·02, 1·06; P < 0·001) and 3·5 % in women (RR 1·03; 95 % CI 1·01, 1·06; P < 0·05) with each 0·1 unit increase of WHR by sex-specific linear dose–response analyses. When the non-linear dose–response analysis was used, a significant non-linear relationship was found (P < 0·001). The dose–response plot showed a slow slope when WHR increased to 0·9, and when WHR values were higher than 0·9, the risk of CVD rose more obvious in men. While in female participants, the curve with an inconspicuous inflexion point reminded that the risk of CVD enhanced continuously with the increase of WHR (Fig. 5).
Fig. 4. Forest plot of association between waist:hip ratio and risk of CVD. ES, effect size.
Fig. 5. Dose–response analysis between waist:hip ratio and risk of CVD men (left) and women (right). Solid line: non-linear dose response. Dotted line: 95 % CI. Straight dotted line: linear dose–response.
Waist:height ratio and CVD
Nine studies involving 187 137 participants and 7557 cases reported sufficient information on WHtR and risk of CVD. Compared the highest category with the lowest category, the summary RR was 1·57 (95 % CI 1·37, 1·79; I 2 = 37·6 %; P < 0·001). In analysis of different sexes, the summary RR was 1·67 (95 % CI 1·40, 1·98; P < 0·05) with a low heterogeneity (I 2 = 16·4 %; P heterogeneity = 0·31) in women and 1·56 (95 % CI 1·25, 1·95; P < 0·05) with a moderate heterogeneity (I 2 = 38·2 %; P heterogeneity = 0·152) in men (Fig. 6). Four studies were available in dose–response analyses. The linear dose–response analysis revealed that the risk of CVD increased by 8·6 % in men (RR 1·08; 95 % CI 1·01, 1·16; P < 0·05) and 6·0 % in women (RR 1·06; 95 % CI 1·03, 1·09; P < 0·001) with each 0·1 unit increase of WHR. There was a non-linear relationship between WHtR and risk of CVD in men (P < 0·05) and women (P < 0·001) with the random effects dose–response analysis. The non-linear dose–response curve revealed that the risk of CVD would rise sharply both in men and women when WHtR over 0·5 (Fig. 7).
Fig. 6. Forest plot of association between waist:height ratio and risk of CVD. ES, effect size.
Fig. 7. Dose–response analysis between waist:height ratio and risk of CVD among men (left) and women (right). Solid line: non-linear dose response. Dotted line: 95 % CI. Straight dotted line: linear dose–response.
Subgroup analysis, publication bias, sensitivity analysis
The subgroup analysis was conducted to explore the potential sources of heterogeneity. Subgroup analyses were carried out by outcome, age, geographic location, follow-up duration, number of participants and quality of articles in male and female (Tables 1–3). Most of the subgroups still kept relevant to CVD risk after subgroup analysis. When subgroup analysis was performed by outcomes, we found that the association is stronger in women than in men in relationship between CHD and WC and WHR. However, perhaps due to the limited studies, the same phenomenon has not been observed in WHtR. In the sensitivity analysis, one single study was removed at a time and the analysis was conducted with the remaining studies to assess whether the result was affected by the excluded study. The result revealed that no individual study affected the pooled effect size (Figs. S1–S3). The funnel plot, Egger’s test and Begg’s test were used to analyse the publication bias. No evidence of publication bias was found in the relationship between CVD and WC and WHtR (Figs. S4 and S6). However, publication bias was observed in study of the relationship between CVD and WHR (P < 0·05), and the funnel plot seemed to be asymmetric (Figs. S5). After using the ‘trim-and-fill’ method to correct the publication bias, the result was still statistically significant (RR 1·36, 95 % CI 1·25, 1·47).
Table 1. Subgroup analyses of waist circumference (WC) and risk of CVD in men and women
(Risk ratios (RR) and 95 % confidence intervals)
MI, myocardial infarction; VTE, venous thromboembolism.
Table 2. Subgroup analyses of waist:hip ratio (WHR) and risk of CVD in men and women
(Risk ratios (RR) and 95 % confidence intervals)
MI, myocardial infarction; VTE, venous thromboembolism.
Table 3. Subgroup analyses of waist:height ratio (WHtR) and risk of CVD in men and women
(Risk ratios (RR) and 95 % confidence intervals)
MI, myocardial infarction; VTE, venous thromboembolism.
Discussion
CVD has been becoming a serious public health burden around the world. Obesity, especially abdominal obesity is closely related to CVD. WC, WHR and WHtR are precise indicators of abdominal obesity, which have advantage of easily available. Previous studies have demonstrated that indicators of abdominal obesity were better than BMI for prediction of many diseases(Reference Radholm, Chalmers and Ohkuma8–Reference Savva, Lamnisos and Kafatos11). Identifying the relationship between these regulable factors and CVD will help reduce the risk of CVD. However, many researchers explored the association between abdominal obesity indices and cardiovascular risk factors, such as hypertension, the metabolic syndrome and type 2 diabetes. A comprehensive meta-analysis to investigate the relationship between WC, WHR, WHtR and incidence of CVD is urgent to be done.
Our meta-analysis comprehensively evaluated the relationship between WC, WHR, WHtR and the risk of CVD in general population with linear and non-linear dose–response models incorporating the data from thirty-one prospective studies. These results showed significant correlations of WC, WHR, WHtR with cardiovascular risk. Compared with the lowest category of WC, WHR and WHtR, the CVD risk of highest group was increased by 43, 43 and 57 %, respectively. The linear dose–response analysis showed the risk of CVD increased by 4·0 and 3·4 % for each 10 cm increase of WC in men and women, respectively. When WHR and WHtR increase by 0·1 units, the incidence of CVD enhanced by 4·0 and 8·6 % in men, and increased by 3·5 and 6·0 % in women, respectively.
Non-linear analysis showed that the risk of CVD rose continuously when WC was over 90 cm, WHR was over 0·9, WHtR was over 0·5 in men, and when WC was over 80 cm, WHtR was over 0·5 in women. It was interesting to note the non-linear dose–response curve with inconspicuous inflexion point in women. This trend showed that the risk of CVD increased continuously with WHR in women. Some epidemiological studies have demonstrated that the risk estimates for WC without hip circumference adjustment would underestimate the risk of CVD(Reference Yusuf, Hawken and Ounpuu45,Reference Canoy, Boekholdt and Wareham46) , which reminded us that large hip circumference may be a protective factor for CVD in women.
The results of non-linear analysis can also help determine the boundary markers for prevention of CVD. Cut-off points recommended by WHO to define abdominal obesity are as follows: WC > 102 cm, WHR > 1·0 for men, and WC > 88 cm, WHR > 0·85 for women. According to our analysis, men should keep their WC below 90 cm and WHR below 0·9 to maintain cardiovascular fitness. As for women, they should keep their WC below 80 cm and WHR should be as small as possible within the normal range. WHtR below 0·5 applies to both men and women.
It is worth mentioning that the existing literature suggested that 0·5 was an appropriate boundary value for predicting abdominal adiposity, cardiovascular risk and all-cause mortality. Research from Jayedi et al. of seventy-two prospective cohort studies suggested that the risk of all-cause mortality increased sharply when WHtR was >0·5(Reference Jayedi, Soltani and Zargar47). Browning et al.(Reference Browning, Hsieh and Ashwell48), Castanheira et al.’s research conducted in 9246 Brazilian adults(Reference Castanheira, Chor and Braga49) and Gibson et al.’s recent data from 4112 adults in England(Reference Gibson and Ashwell50) had shown that 0·5 was a suitable cut-off point for predicting cardiovascular risk. Several other studies also supported this point of view(Reference Garnett, Baur and Cowell51,Reference Ashwell and Gibson52) . Our research demonstrated that the risk of CVD increased sharply when WHtR was over 0·5 through analysis of 187 137 participants from different regions and races. This strongly supported the simple public health message: Keep your WC to less than half your height.
Many scientists have been studying how abdominal obesity up-regulates the risk of CVD. Some researchers believed that increased abdominal fat is a marker of more ectopic fat in some organs, including liver and heart(Reference Smith53). Fat tissue can be divided into subcutaneous fat tissues and visceral adipose tissue according to their distribution. When fat accumulates in vital organs such as liver and heart, it is likely to be closely associated with an increased CVD risk. Therefore, many researchers have put forward ideas that the essential role of visceral fat in the development of CVD should be taken into account(Reference Shimabukuro54). The expansion of adipose tissue would induce the occurrence of adipokine secretion and function dysregulation, which play a fundamental role in development of inflammatory disorders, cardiometabolic disorders and vascular disorders, ultimately leading to a variety of cardiovascular events(Reference Harwood55). In addition, studies have found that pericardial fat and visceral adipose tissue were associated with CVD after age and sex adjustment for quantitative volume of adipose tissue by chest multidetector computed tomography(Reference Mahabadi, Massaro and Rosito56).
In addition, many researchers take WHtR as a preferable indicator in the prediction of CVD. The study of Castanheira et al., a longitudinal study of Brazilian(Reference Castanheira, Chor and Braga49), revealed that WHtR was better than other adiposity for predicting CVD with good discriminatory power in Brazilian adults. Ashwell’s meta-analysis also supported this conclusion, which was conducted in 123 231 men and 182 620 women with hypertension, type 2 diabetes, dyslipidaemia, the metabolic syndrome and general CVD(Reference Ashwell, Gunn and Gibson9). Our study included a larger sample size of healthy adults which confirmed this conclusion to some extent.
In conclusion, our research showed that, as good indicators of abdominal obesity, WC, WHR and WHtR were significantly associated with the incidence of CVD. WHtR may be better than WC and WHR, and 0·5 could be an appropriate boundary value for the prediction of CVD.
This study has several strengths, we systematically assessed the linear and non-linear dose–response relation between WC, WHR, WHtR and the risk of CVD, and then we found all of these three indicators were closely related to the risk of CVD. A total of approximately 670 000 participants and 25 000 cases were included in our study, which could help minimise potential bias preferably. To eliminate the interference of sex, we conducted subgroup analyses of each sex based on the study and participant characteristics.
There are still some limitations in this study. Although the estimates in our analyses were derived from maximally adjusted models to reduce potential confounding factors, the adjusted factors for each study are not identical and some confounders still exist. This is a possible source of heterogeneity. Some of studies included in our analysis collected information via self-reporting questionnaires, which might produce errors in the measurement of WC, WHR and WHtR. The publication bias was found in the analysis of relationship between WHR and CVD risk, while the result was still statistically significant when ‘trim-and-fill’ was applied to correct the bias. Therefore, the results were basically reliable. Furthermore, most available studies have focused on the degree of adiposity as important risk factors in the development of CVD, but the effect of the duration of obesity was ignored. The consequences of exposure to higher levels and longer duration of obesity should be taken into account in future studies.
Conclusion
According to the analysis of existing evidence, we found that the risk of CVD is related to abdominal obesity. The risk of CVD rose continually with the increase of WC, WHR and WHtR when they exceeded a certain range. Keep your WC to less than half your height could help reduce the risk of CVD. As sample and useful indicators of abdominal obesity, WC, WHR and WHtR, especially WHtR, are worth to popularise in future study and clinical application to help prevent CVD.
Acknowledgements
This work was supported by grants from the National Science Foundation of China (grant no. 81903314), Key Scientific Research Project of Henan Province (grant no. 17A330006) and China Postdoctoral Science Foundation (grant no. 2017M622379).
R. X. designed the study, concept and performed the data analysis. R. X., Q. L. and Y. G. were involved in the data collection and analysis. R. X. wrote the first draft of the manuscript. All authors critically revised the manuscript and gave final approval and agreed to be accountable for all aspects of the work to ensure the integrity and accuracy. All the work was done under S. Z.’s direction.
The authors do not have any conflicts of interest to declare.
Supplementary material
For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114521000064
