Discriminatory Capacity of Anthropometric Indices for Cardiovascular Disease in Adults: A Systematic Review and Meta-Analysis

Introduction Obesity is one of the main risk factors for cardiovascular disease (CVD) and cardiometabolic disease (CMD). Many studies have developed cutoff points of anthropometric indices for predicting these diseases. The aim of this systematic review was to differentiate the screening potential of body mass index (BMI), waist circumference (WC), and waist-to-hip ratio (WHR) for adult CVD risk. Methods We used relevant key words to search electronic databases to identify studies published up to 2019 that used receiver operating characteristic (ROC) curves for assessing the cut-off points of anthropometric indices. We used a random-effects model to pool study results and assessed between-study heterogeneity by using the I 2 statistic and Cochran’s Q test. Results This meta-analysis included 38 cross-sectional and 2 cohort studies with 105 to 137,256 participants aged 18 or older. The pooled area under the ROC curve (AUC) value for BMI was 0.66 (95% CI, 0.63–0.69) in both men and women. The pooled AUC values for WC were 0.69 (95% CI, 0.67–0.70) in men and 0.69 (95% CI, 0.64–0.74) in women, and the pooled AUC values for WHR were 0.69 (95% CI, 0.66–0.73) in men and 0.71 (95% CI, 0.68–0.73) in women. Conclusion Our findings indicated a slight difference between AUC values of these anthropometric indices. However, indices of abdominal obesity, especially WHR, can better predict CVD occurrence.


Introduction
Although many factors for cardiovascular disease (CVD) have been identified, the number of deaths from CVD worldwide rose from 12.6 million to 17.6 million between 1990 and 2016 (1,2). CVD is the most common cause of death in both developed and developing countries; the CVD mortality rate was more than 900,000 in the United States in 2016 (2,3).
Obesity, especially abdominal obesity, is a modifiable CVD risk factor that is increasingly prevalent worldwide (4). Abdominal obesity refers to the accumulation of fat in the central area of the body, which can lead to adverse effects such as hypertension, in-sulin resistance, and hyperlipidemia (5,6). The most common anthropometric indices used to screen for obesity and overweight are body mass index (BMI, weight in kg/height in m 2 ), waist circumference (WC), and waist-to-hip ratio (WHR) (2,7,8). BMI is a simple indicator associated with an increased risk of CVD, although it may not reflect variations in body fat distribution (9). Because of its simplicity, usability, and availability, BMI is the most common method of obesity assessment (10). WC and WHR are also good indicators of abdominal obesity and, similar to BMI, can predict cardiometabolic disorders (9,11).
The World Health Organization (WHO) recommends a BMI cutoff point of 25.0 for overweight and 30.0 for obesity and a WC of 102 cm (40 inches) in men and 88 cm (35 inches) in women as cut-off points for abdominal obesity (12). Because of the increasing prevalence of obesity worldwide, many studies have aimed to determine optimal cut-off points of anthropometric indices (7,13,14). Furthermore, because of racial/ethnic differences in body composition, WHO encourages researchers to conduct studies to determine the cut-off points of anthropometric indices in different populations (15). However, these racial/ethnic differences and differences in study design have led to variations in findings as to which indices better predict these diseases (16).
Despite the many studies that have assessed optimal cut-off points of anthropometric indices for predicting CVD, there is no study that summarizes these findings. Moreover, no comprehensive information is available on which index -BMI, WC, or WHRbetter predicts CVD. Therefore, we conducted a systematic review and meta-analysis of the studies that analyzed these 3 indices to assess their effectiveness in predicting CVD.

Methods
We used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) as the basis of our systematic review and meta-analysis (17). The study protocol was registered in the database of the International Prospective Register of Systematic Reviews (PROSPERO) in June 2019 (registration no. CRD42019121324).

Search strategy
We searched international databases including Web of Science, Medline via PubMed, Scopus, Cochrane Library, ProQuest, and Google Scholar in July 2019. We also searched national databases in Iran, including Magiran and SID (Science Information Database). We did not limit our search to a specific timeframe. Additional studies were identified from manual reference checks of selected studies. We used a sensitive search strategy to retrieve more relevant studies.
We used Boolean operators (ie, AND, OR, and NOT) to perform the search. We used AND to search both common terms, OR to find information that included either search term, and NOT for terms that we did not want to retrieve. We used parentheses to combine the search terms by outcome, exposure, and population categories. We used quotation marks to search for exact terms or expressions.

Eligibility criteria and data extraction
In accordance with the PECO (Population, Exposure, Comparator, and Outcomes) framework, we included all original articles from cross-sectional and prospective cohort studies that examined the optimal cut-off points of BMI, WC, and WHR for predicting CVD, regardless of limitations in age, sex, language, race/ethnicity, and publication year. The study population included healthy adults (aged ≥18 y). Studies were included regardless of differences in measurement methods. Studies on children, adolescents, or a subgroup of patients (eg, cancer, HIV, pregnancy) were excluded. Two reviewers appraised the studies independently on the basis of inclusion criteria.
Data for the included articles were summarized as first author; year of publication; participants' age, sex, and nationality; sample size; study design; cut-off points (BMI, WC, and WHR); area under the receiver operating characteristic (ROC) curve (AUC) (95% CI); and sensitivity and specificity in prespecified data extraction form in Excel (Microsoft Corporation).

Outcomes
The outcomes of interest were CVD and cardiometabolic disease (CMD). CVD was defined as conditions that involve narrowed or blocked blood vessels that can lead to ischemic heart disease, chest pain (angina), myocardial infarction, and stroke. CMD was defined as a condition in which there is a high possibility of developing atherosclerotic CVD and diabetes mellitus (18). question about appropriate design; 2) sampling method and adequate sample size; 3) place and date of the research; 4) expression of study type; 5) a question about acceptable response rate; 6) full description of inclusion and exclusion criteria and demographic characteristics; and 7) method of measuring the health outcome (19).

Exposure cut-off point selection
The search of the included studies indicated that reporting of exposure cut-off points was based on different methods by the researchers: 1) optimal cut-off points, or those that were chosen to maximize sensitivity and specificity of the indices; and 2) studies that reported the AUC and associated 95% CIs. The AUC is commonly used for assessing the discriminative ability of predictive and prognostic models to discriminate between individuals who will or will not develop the disease. The AUC is used to compare the accuracy of a test, where a greater area indicates that the test is more accurate (20,21). An AUC of less than 0.60 was considered to have poor diagnostic performance (22).

Statistical analysis
The heterogeneity of the studies was assessed by using the Cochran Q test (with significance of P < .10 because of the low power of the test) and the I 2 statistic (22). We used a randomeffects model with the inverse-variance method and developed forest plots to describe the results and calculate the point estimations and 95% CIs. Forest plots are used to depict the included studies, demonstrate the differences between studies, and provide estimates of overall results (23). We used subgroup analysis to explore potential sources of heterogeneity, and we used Begg's and Egger's tests to investigate potential publication bias. We used Stata software version 14.2 (Stata Corp LLC). Significance was set at P < .05.

Study selection
Our search yielded 2,457 records; after duplicate articles were eliminated, 1,588 records remained. We then excluded 1,356 records because the articles were deemed irrelevant on the basis of their titles or abstracts, leaving 232 studies remaining for full-text analysis. In this step, 194 studies were excluded for the following reasons: no relevant outcome measure or data available (n = 146); studies were conducted on a subpopulation (n = 9); full-text art-icle not available in English (n = 10); article was a systematic review or meta-analysis (n = 5); article did not report optimal cut-off points, AUC, or sensitivity and specificity (n = 19); article was a conference abstract (n = 2); or analysis not conducted in adults (n = 3). In total, we identified 38 qualifying studies that were included in the meta-analysis ( Figure 1).

Study characteristics
Of the 38 articles, 36 were cross-sectional studies and 2 were cohort studies (Table). Studies were conducted from 1996 to 2016 in 16 different countries. The age limit for inclusion in each of the individual studies ranged from 18 to 90 years. The study population size ranged from 105 to 137,256 participants.

Results of the meta-analysis
We created forest plots of AUC scores based on BMI, WC, and WHR for CVD and CMD risk in men and women. Based on the random-effects model, the pooled AUC value for BMI was 0.66 (95% CI, 0.63-0.69) both in men and women (    The pooled sensitivity value for BMI with CVD or CMD was 0.62 (95% CI, 0.58-0.65) in men and 0.62 (95% CI, 0.58-0.66) in women, and the pooled sensitivity value for WC in men was 0.68 (95% CI, 0.66-0.70) and in women was 0.67 (95% CI, 0.64-0.69). The pooled sensitivity value for WHR was 0.66 (95% CI, 0.64-0.69) in men and 0.66 (95% CI, 0.62-0.69) in women.
We conducted 4 subgroup analyses to address the effect of the sex, study location, year of publication, and quality of included studies as potential sources of the observed heterogeneity. We found that sex was one source. Heterogeneity was still appreciable for all subgroups, but the AUC, sensitivity, and specificity differences in values between subgroups were not significant. The results of Begg's test for CVD based on BMI, WC, and WHR was not significant, so we determined that there was no evidence of publication bias.

Meta-regression
The results of the random-effects meta-regression analysis indicated that year of study (coefficient = −0.03; P = .34), location of study (coefficient = 0.03; P = . 16), and year of publication (coefficient = −0.03; P = .34) were not significant moderators of the observed heterogeneity. However, we found that type of study was a potential a source of heterogeneity (coefficient = −0.14, P = .04).

Discussion
Our study is the first to summarize findings on the ability of anthropometric indices' cut-off points to predict CVD, using 38 cross-sectional and prospective studies with 105 to 137,256 participants. Our findings showed that all examined anthropometric indices have moderate power in CVD and CMD screening, for which the AUC values were significantly greater than 0.6. However, WC and WHR better predicted CVD than did BMI.
Obesity is a risk factor for CVD development. Traditionally, BMI is the most commonly used index for assessing overweight and obesity (9), but BMI is a predictor of overall obesity without consideration of sex (25). Because it is known that type of fat distribution (android or gynoid) has an effect on CVD pathogenesis, BMI cannot accurately represent central adiposity (25,26). Furthermore, many people who present with abdominal obesity also have a low BMI (24).
Increased WC is associated with increased adipocytes in this area. In obesity, adipocytes grow, enlarge, and secrete inflammatory cytokines, such as tumor necrosis factor α, interleukin-6, and highsensitivity C-reactive protein (27). Excess adipose tissue as an inflammatory tissue can lead to chronic inflammation in the body, which has an adverse effect on the pathophysiology of atherosclerosis and CVDs (27,28). Furthermore, high body fat causes leptin resistance and inhibits lipolysis by producing matrix metalloproteinase-2 (29,30). Therefore, the ability of WC and WHR to better predict CVD can be explained by their assessment of abdominal fat, with its role in secreting inflammatory cytokines and inducing leptin resistance.
Many of the studies we reviewed showed that indices of abdominal obesity can better predict CVDs (31-39) and CMD (7,16,40,41 (16) observed that WHR is a better predictor for CVDs and CMD than are other evaluated indices. Results from a meta-analysis in 2011 on 82,864 British participants from 9 cohort studies showed that indices of abdominal obesity, including WC and WHR, were related to CVD mortality and that BMI had no relation to CVD mortality (43). Another meta-analysis on more than 88,000 parti-cipants in 2008 by Lee et al supported the conclusion that indices of abdominal obesity are better predictors of CVD risk factors compared with BMI (44). Also, a meta-analysis in 2012 by van Dijk et al on 20 articles with 45,757 participants found that indices of abdominal obesity, especially WC, are more strongly predictive of CVD risk factors (45). Evidence from a meta-analysis and systematic review by Cao et al on 26 case-control and trial studies determined that WHR can predict the occurrence of myocardial infarction in both sexes (46).
Growing evidence shows that higher energy intake results in stored fat in the central area of the body (47), and excessive fat accumulation is linked with ectopic fat deposition in the liver, pancreas, and skeletal muscle. This ectopic fat accumulation can increase risk of developing features of diabetes, dyslipidemia, metabolic syndrome, CVDs, and overall CMDs (48-50). Increased hip circumference indicates an increase in fat accumulation in the gluteal muscles and lower limbs, which is associated with decreased physical activity, and this may be a potential risk factor for CMDs (46,47).
A strength of this review was the large number of included studies. The study had limitations. Most studies were conducted in Asian countries, with few studies on other continents. Another limitation was that some studies reported results based on AUC and some with sensitivity and specificity; it was not possible to combine these 2 values, so we had to divide the articles into 2 groups and analyze them separately.
In conclusion, this systematic review attempted to summarize the evidence on anthropometric indices cut-off points for predicting CVDs, and which indices better predict these diseases. On the basis of our findings, all 3 indicators are good screening tools for predicting CVD. However, indices of abdominal obesity, especially WHR, can better predict CVD occurrence. Future studies should include children and adolescents in the study population.    Table   Table. Characteristics of Studies Included in a Systematic Review and Meta-Analysis of the Discriminatory Capacity of Anthropometric Indices for Determining Risk for Cardiovascular Disease, 2020