Volume 3 | Issue 2 | 2016-09-09

Validation of a Novel Visceral Adiposity Prediction Tool and Comparison to Existing Adiposity Indices

Pandora Wander

Abstract

Background/objectives: Visceral adiposity measured by computed tomography (CT) as intra-abdominal fat area (IAFA) predicts metabolic diseases. Measuring IAFA by CT, however, is not practical. We developed a new index (Visceral Adiposity Estimate VAE) for IAFA estimation, and validated its performance. Subjects/methods: We recorded age, BMI, waist circumference (WC), fasting lipids, glucose, and smoking status for 622 participants in the Japanese-American Community Diabetes Study (JACDS). IAFA was measured with a single CT slice at the umbilicus. We used stepwise linear regression and regressed IAFA on commonly available laboratory and demographic variables in sex-stratified models. We termed this estimate the VAE. For men, variables included were age, WC, HDL, and glucose; for women, age, BMI, WC, HDL, LDL, and glucose. We compared area under receiver-operating characteristic curves (AUROC) between VAE, the Visceral Adiposity Index (VAI: a measure incorporating BMI, WC, triglycerides, and HDL), and WC for IAFA dichotomized at the median and 75 th percentile. Results were validated in 1645 Japanese participants in the Ohtori Study. Results: Participant characteristics were: JACDS, mean age 53.7, 55% male; Ohtori, mean age 50.4, 63% male. In JACDS, VAE predicted above-median IAFA (men: AUROC = 0.88; women: AUROC = 0.94) better than VAI and WC for men (VAI AUROC = 0.74, p < 0.001; WC AUROC = 0.81, p < 0.001) and women (VAI AUROC = 0.79, p < 0.001; WC AUROC = 0.88, p = 0.001). At the 75 th percentile, VAE predicted IAFA (men: AUROC = 0.88; women: AUROC = 0.93) better than VAI and WC for men (VAI AUROC = 0.74, p < 0.001; WC AUROC = 0.84, p = 0.004) and women (VAI AUROC = 0.78, p < 0.001; WC AUROC = 0.88, p = 0.030). Results were similar for Ohtori. Conclusions: VAE predicted IAFA better than VAI or WC in derivation and validation datasets of Japanese-American and Japanese populations.

Keywords: visceral adiposity, visceral adiposity index, abdominal obesity, waist circumference, anthropometrics

Full text

Excess visceral adipose tissue (VAT) is a major risk factor for metabolic disease in Japanese Americans [1-6] . Surrogate measures of visceral adiposity including BMI and waist circumference (WC) are routinely used [7] in both clinical practice and epidemiologic research as direct measurement of VAT is often prohibitively expensive or logistically difficult to obtain. In 2010, Amato et al. proposed a measure that used laboratory and anthropometric data to estimate visceral adiposity dysfunction associated with cardio metabolic risk, called the visceral adiposity index (VAI) [8] . Since then, VAI has been associated with adverse metabolic phenotypes in Asian and non-Asian populations, including non-alcoholic fatty liver disease in Italians [9] , insulin resistance in Korean women with polycystic ovarian syndrome [10] and type 2 diabetes in Iranian and Chinese cohorts [11-13] . VAI has also been proposed as a surrogate for visceral fat quantity; however, in the handful of small studies that have been conducted, the strength of the association of VAI with intra-abdominal fat area is variable [10,14,15] . Given the varying association of VAI with measured visceral fat quantity, we examined whether an alternate adiposity surrogate might more closely parallel quantity of intra-abdominal fat area. We hypothesized that a new adiposity surrogate developed using regression methods and readily available clinical and anthropometric data could predict computed tomography (CT)-measured intra-abdominal fat area better than existing visceral adiposity surrogates such as VAI and WC. Materials/ Subjects & Methods Study design and setting We measured demographic, metabolic, and body composition variables in a derivation dataset of 622 Japanese- American participants. Study participants were taken from the Japanese American Community Diabetes Study (JACDS), a community-based cohort of second- (Nisei) and third-generation (Sansei) Japanese Americans of 100% Japanese ancestry who were representative of Japanese-American residents of King County, Washington State, in age, residential distribution, and parental immigration pattern. Selection and recruitment have been described previously [16] . Briefly, JACDS was designed to investigate risk factors for and prevalence of type 2 diabetes among Japanese Americans. Individuals were recruited between 1984 and 1992. A comprehensive mailing list and telephone directory that included almost 95% of the Japanese-American population was used to identify and contact potential study subjects. They were eligible to participate if they were 34 years old and of 100% Japanese ancestry by self-report. Of an initial sample of 658 individuals, 36 were missing data on covariates, leaving an analytic sample of 622 participants. For the validation dataset, we used data from 1645 participants in the Ohtori Study who had complete data on all covariates. The Ohtori Study prospectively follows Japanese men and women to examine relationships between body fat distribution and cardiometabolic diseases [17] . Participants who visited the Ohtori Health Promotion Center between December 2005 and April 2010 for a medical checkup were recruited for the Ohtori study. For the current analysis, eligible participants consisted of 1645 individuals with complete data on all covariates. The study received approval from the University of Washington Human Subjects Division and the Human Subjects Review Committee at Osaka City University, and all participants provided written informed consent. Data collection Evaluations were done at the General Clinical Research Center at the University of Washington, Seattle, and the Ohtori Health Promotion Center, in Sakai, Japan. For studies, information on age, sex, and current and former tobacco use was obtained by interview. Trained staff measured height, weight, and WC. Weight was measured using a digital scale after shoes and outer clothing was removed. BMI was defined as weight in kg divided by height in meters squared. In JACDS, WC (cm) was measured twice using a flexible tape measure and averaged, whereas in the Ohtori study, it was measured once. In male JACDS participants, the measurement was obtained at the midpoint between the iliac crest and the inferior border of the lowest rib. In female JACDS participants, the minimal circumference between iliac crest and lowest rib was used. In Ohtori, WC was measured at the level of the umbilicus for all participants. Serum glucose, triglycerides, HDL, and LDL were measured in the fasting state and reported in mg/dL. All specimens were drawn after at least a 10-hour fast. Using the approach and coefficients described by Amato et al., we calculated sex-specific VAI values for all participants, using BMI, WC, triglycerides, and HDL [8] . In that study, VAI was defined as (WC/(39.68 + (1.88 * BMI))) * (triglycerides/1.03) * (1.31/HDL) for men, and (WC/(36.58 + (1.89 * BMI))) * (triglycerides/0.81) * (1.52/HDL) for women, with both triglycerides and HDL specified in mmol/L. In both cohorts, to estimate intra-abdominal fat area, single (1 cm) CT-scan slices were obtained at the level of the umbilicus. Scans were analyzed using density contour software. Areas corresponding to a density of -250 to -50 Hounsfield units for JACDS and -200 to -50 Hounsfield units for the Ohtori study were classified as adipose tissue [18] . Intra-abdominal fat area was defined as fat within the confines of the transversal is fascia and reported in cm&sup2;. In JACDS, intra-observer variability for multiple measurements by a single observer of a single CT scan ranged from 0.2% to 1.4%. Statistical analysis We used number (%) and mean (standard deviation) for categorical and normally distributed continuous variables, respectively, to describe study population characteristics, both overall and stratified by sex. For fasting triglycerides, we reported median (inter-quartile range), as the distribution was skewed. We examined relationships of three visceral adiposity surrogates with measured intra-abdominal fat area using both continuous and categorical measures. First, we created a novel visceral adiposity surrogate. We used backward stepwise linear regression to fit models in the training cohort using easy-to-obtain demographic, metabolic, and anthropometric variables (age, BMI, WC, HDL, LDL, triglycerides, fasting glucose, and smoking status) to predict intra-abdominal fat area. A p-value 0.05 allowed entry into the model, while a p-value of 0.1 or greater prompted removal. Because there was evidence of a multiplicative first-order interaction of sex with WC, we chose to perform analyses stratified by sex. We also fit linear regression models examining the association of two other popular visceral adiposity surrogates (VAI and WC) with intra-abdominal fat area as the dependent variable. We fit linear regression models looking for evidence of nonlinearity in associations of metabolic variables and body composition with intra-abdominal fat area using the Stata command fracpoly . We also performed model diagnostics to identify influential outliers and for evidence of violations of model assumptions [19] . Secondly, we dichotomized intra-abdominal fat area at the median and 75 th percentile values. We then plotted receiver- operating characteristic (ROC) curves comparing performance of the three visceral adiposity surrogates in discriminating low vs. high categories of intra-abdominal fat area. We used median and 75 th percentile cutpoints taken from the training dataset for all analyses (median: 100 cm&sup2; [men] and 70 cm&sup2; [women]; 75 th percentile: 140 cm&sup2; [men] and 100 cm&sup2; [women]). We compared the area under the ROC curves (AUROC) of VAI and WC to VAE using chi-squared tests [20] . We then used the coefficients from the derivation dataset to predict intra-abdominal fat area among Ohtori participants. We regressed the linear predictor against intra-abdominal fat area to obtain R&sup2; values, plotted ROC curves, and used Bonferroni-corrected chi-squared tests to compare discrimination of low vs. high intra-abdominal fat area in the validation cohort. For both cohorts, we reported R&sup2; from linear regression models as well as area under the ROC curves and 95% CI. As a sensitivity analysis, we repeated the analyses comparing AUROC for VAE, VAI, and WC for the dichotomized intra-abdominal fat area among Ohtori participants, using median and 75 th percentile cut points from that dataset, instead of from the training dataset (median: men 77.9 cm&sup2;, women 35.4 cm&sup2;; 75 th percentile: men 107.3 cm&sup2;, women 63.4 cm&sup2;). Lastly, we compared performance of VAE to a larger list of adiposity surrogates in the discrimination of low vs. high intra-abdominal fat area. For these results, shown in Supplementary Table 2, we performed analyses in two ways by using the coefficients of published indices to generate predicted VAT when possible, and then by fitting regression models to our data using the variables and transformations of variables as reported in published indices. This extended list of adiposity surrogates was composed of fitted VAI (a version of VAI obtained by repeating the steps Amato et al. used in our dataset to obtain coefficients specific for our Japanese-American population) [8] ; a method derived by Despr s et al. (1991) incorporating age and WC, using both the original coefficients he derived and coefficients specific to our dataset [21] ; a method derived by Kahn et al. (2003) [22] incorporating dichotomous measures of excess WC and excess triglyceride levels; the lipid accumulation product (LAP), which includes waist circumference and triglycerides [23] ; and the deep-abdominal adiposity tissue (DAAT) index, which includes weight, waist circumference, and BMI [24] . For all analyses, a p-value of < 0.05 was considered statistically significant. For primary results, we reported Bonferroni- corrected p-values. Analyses were performed with Stata (version 12.0; StataCorp, College Station, TX). Results Table 1 shows participant characteristics overall and stratified by sex. Coefficients and p-values from sex-specific linear regression models testing associations of metabolic risk factors with intra-abdominal fat area in the derivation cohort are shown in Table 2. All the variables shown in Table 1 with the exception of intra-abdominal fat area and diabetes status were considered in the multivariable models described in Table 2. Using the stepwise methods described above, BMI, LDL, triglycerides, and smoking status were dropped from the model in men, leaving age, WC, HDL, and fasting glucose. In the model for women, triglycerides and smoking status were dropped from the model, leaving age, BMI, WC, HDL, LDL, and fasting glucose. A test of regression assumptions found that for the model in women, the linear form of intra-abdominal fat area violated regression assumptions, while the square root transformation did not; however, nearly identical results were obtained in Tables 2 and Table 3 with either the linear or square root transformation of intra-abdominal fat area as the dependent variable, so the linear version was used to permit easier interpretation of our findings. An examination of best fitting transformations of the independent variables using fracpoly revealed that the best fit was obtained with the linear versions of these covariates as shown in Table 2. Model diagnostics did not identify influential outliers, and there was no further evidence of violations of model assumptions. The final versions of our models are: for men, predicted IAFA = (1.28 * age) + (4.12 * WC) (0.53 * HDL) + 0.14 * fasting glucose) 319.5. For women, predicted IAFA = (1.26 * age) + (1.89 * BMI) + (2.16 + WC) (0.43 * HDL) + (0.11 * LDL) + (0.18 * fasting glucose) 207.2. Results comparing the models shown in Table 2 with other adiposity surrogates in prediction of intra-abdominal fat area are shown in Table 3. In the derivation dataset, VAE exhibited a higher R&sup2; than VAI or WC in multiple linear regression models with continuous intra-abdominal fat area as the dependent variable for both sexes. The same was true for classification of intra-abdominal fat area above the median or 75 th percentile. Similar results were seen in the validation dataset for the abilities of the adiposity surrogates to predict intra-abdominal fat area either continuously or categorically (Table 3). In the validation dataset, VAE showed the best classification of intra-abdominal fat area above the median (VAE men: AUROC = 0.86; women: AUROC = 0.92) compared to VAI for men (AUROC = 0.72, p < 0.001) and VAI and WC for women (VAI AUROC = 0.79, p < 0.001; WC AUROC = 0.88, p = 0.001). For men in the validation population, VAE and WC were not significantly different at the median cut point (WC: AUROC = 0.85, p = 0.084). At the 75 th percentile cut point, VAE classified intra-abdominal fat area (men: AUROC = 0.90; women: AUROC = 0.95) better than VAI and WC for men (VAI AUROC = 0.74, p < 0.001; WC AUROC = 0.88, p = 0.0025) and women (VAI AUROC = 0.79, p < 0.001; WC AUROC = 0.92, p = 0.030). Results were similar when median and 75 th percentile cutpoints specific to Ohtori were used (Supplementary Table 1). In comparisons among the extended list of adiposity surrogates (Supplementary Table 2) the AUROC for VAE was significantly greater than every other surrogate for men at the median cutpoint. It was also significantly greater than every adiposity surrogate at the 75 th percentile with the exception of the Despr s adiposity-tissue estimate calculated by fitting the regression model to our data. For women, the AUROC for VAE was significantly greater than all other surrogates at the 75 th percentile and at the median with the exception of the median Despr s adiposity-tissue estimate calculated by fitting the regression model to our data. Table 1: Baseline characteristics of study participants in the derivation (Japanese American Community Diabetes Study) and validation (Ohtori) cohorts. Japanese American Community Diabetes Study Ohtori Study Overall Men Women Overall Men Women n = 622 n = 344 n = 278 n = 1645 n = 1033 n = 612 Age, years (SD) 53.7 (11.8) 54.2 (11.4) 53.1 (12.4) 50.4 (9.9) 50.3 (9.6) 50.5 (10.5) Sex, % male (n) 55 (344) -- -- 63 (1033) --- -- BMI, kg/m&sup2; (SD) 24.5 (3.3) 25.5 (3.0) 23.2 (3.2) 23.1 (3.2) 23.9 (3.0) 21.8 (3.1) Waist circumference, cm (SD)* 83.3 (10.5) 88.8 (8.2) 76.5 (9.0) 83.0 (9.2) 85.8 (8.3) 78.3 (8.8) HDL, mg/dL (SD) 57 (17) 50 (13) 65 (16) 58 (13) 54 (12) 65 (13) LDL, mg/dL (SD) 141 (37) 146 (38) 135 (34) 124 (31) 124 (31) 124 (32) Median triglyceride level, mg/dL (IQR) 118 (79-167) 133 (95-207) 95 (67-135) 91 (62-135) 109 (72-157) 72 (52-99) Fasting glucose level, mg/dL (SD) 106 (37) 114 (43) 97 (25) 101 (20) 105 (23) 95 (10) Smoking status Current smoker, % (n) 13 (80) 15 (50) 11 (30) 31 (504) 40 (416) 15 (88) Former smoker, % (n) 42 (261) 58 (198) 23 (63) 21 (343) 29 (296) 8 (47) Diabetes at baseline, % yes (n) 20 (127) 26 (88) 14 (39) 7 (108) 9 (93) 2.5 (15) Intraabdominal fat area (cm 2 ) 90.6 (53.3) 105.3 (55.3) 72.4 (44.7) 67.4 (43.6) 81.1 (43.1) 44.2 (33.6) * For women, narrowest measurement of waist circumference was used Abbreviations: SD (standard deviation) Table 2: Multiple regression model testing associations of metabolic risk factors with intra-abdominal fat area (cm&sup2;) in the derivation cohort. Men Women n = 344 n = 278 p-value p-value Age, years 1.28 < 0.001 1.26 < 0.001 Body mass index, kg/m 2 -- -- 1.89 0.063 Waist circumference, cm 4.12 < 0.001 2.16 < 0.001 HDL, mg/dL -0.53 0.001 -0.43 < 0.001 LDL, mg/dL -- -- 0.11 0.027 Triglyceride level, mg/dL -- -- -- -- Fasting glucose level, mg/dL 0.14 0.003 0.18 0.027 Smoking status % current (n) -- -- -- -- % former (n) -- -- -- -- Constant term -319.50 < 0.001 -207.2 < 0.001 On first-order interaction testing, there was a significant interaction of waist circumference with sex in the full model. We therefore chose to present sex-stratified results. We built reverse stepwise prediction models using all covariates above. P-value for entry was set at 0.05. P-value for removal was set at 0.1. For men, BMI, LDL, triglycerides, and smoking status were dropped from the model. For women, triglycerides and smoking status were dropped from the model. Table 3: Model R&sup2; from linear regression models predicting continuous intra-abdominal fat area, and AUROC and 95% CI for prediction of visceral adiposity category at median and 75 th percentile cutpoints by logistic regression models. Japanese American Community Diabetes Study Ohtori Study Men Women Men Women n = 344 n = 278 n = 1033 n = 612 R&sup2; AUROC (95% CI) p-value R&sup2; AUROC (95% CI) p-value R&sup2; AUROC (95% CI) p-value R&sup2; AUROC (95% CI) p-value Prediction of visceral adiposity, linear model VAE 0.59 0.72 0.59 0.64 VAI 0.08 0.14 0.05 0.23 WC 0.48 0.56 0.55 0.56 Prediction of visceral adiposity category at median cutpoints* VAE 0.88 (0.84 - 0.91) - 0.94 (0.92 - 0.97) - 0.86 (0.84 - 0.88) - 0.92 (0.89 - 0.94) - VAI 0.74 (0.69 - 0.79) < 0.001 0.79 (0.73 - 0.84) < 0.001 0.72 (0.68 - 0.75) < 0.001 0.79 (0.75 - 0.83) < 0.001 WC 0.81 (0.77 - 0.86) < 0.001 0.88 (0.84 - 0.92) < 0.001 0.85 (0.82 - 0.87) 0.084 0.88 (0.85 - 0.91) < 0.001 Prediction of visceral adiposity category at 75 th percentile cutpoints VAE 0.88 (0.84 - 0.92) - 0.93 (0.90 - 0.96) - 0.90 (0.88 - 0.93) - 0.95 (0.92 - 0.97) - VAI 0.73 (0.67 - 0.79) < 0.001 0.78 (0.72 - 0.83) < 0.001 0.74 (0.69 - 0.79) < 0.001 0.79 (0.72 - 0.86) < 0.001 WC 0.84 (0.79 - 0.88) 0.004 0.88 (0.84 - 0.92) 0.001 0.88 (0.84 - 0.91) 0.0025 0.92 (0.87 - 0.96) 0.030 Abbreviations: AUROC (area under the receiver-operating characteristics curve), VAE (visceral adiposity estimate), VAI (visceral adiposity index), WC (waist circumference). *The median of intra-abdominal fat area was 100 cm&sup2; for men and 70 cm&sup2; for women. For participants in the Ohtori Study, we used the same cutpoints. The 75 th percentile of intra-abdominal fat area was 140 cm&sup2; for men and 100 cm&sup2; for women. For participants in the Ohtori Study, we used the same cutpoints. Discussion In this population-based analysis of Japanese-American and Japanese participants, we developed and validated a novel adiposity surrogate, the Visceral Adiposity Estimate, to predict visceral adiposity in men and women, and demonstrated that it performed better than WC and VAI in our primary analysis, and as well as or better than other less commonly used or more complex adiposity surrogates in a secondary analysis. The VAE showed a higher correlation with and greater discriminatory ability than existing surrogate measures used to measure visceral adiposity. It relies on readily available clinical and laboratory measurements and can easily be calculated for men and women using the regression coefficients and constants in Table 2. Furthermore, we have demonstrated the applicability of this instrument to not only the population in which it was developed, but also a separate Japanese population. Whether the VAE would be similarly accurate in other populations is possible and should be the subject of future research. We are unaware of previous studies formally estimating the ability of VAI or WC to predict intra-abdominal fat area in Japanese Americans specifically; however, previous studies have quantified relationships of VAI and WC with intra-abdominal fat area in other Asian and non-Asian populations [10,14,15] . Different analytic methods make direct comparisons across the studies somewhat difficult. In a study of 99 overweight post-menopausal Canadian women enrolled in a weight reduction randomized controlled trial, baseline VAI was correlated with CT-measured VAT in cross-sectional analysis of a generally similar magnitude to what we observed in our Japanese-American men (r = 0.284) [15] . A ROC analysis of the ability of the VAI to classify these study participants into the lowest or highest tertile of CT-measured VAT demonstrated an AUROC of 0.61 that was of significantly lower magnitude than the AUROCs for BMI (0.84) or WC (0.86). In an Australian study of 39 men and women with non-alcoholic fatty liver disease, VAI was modestly correlated with intra-abdominal fat area on MRI (r = 0.39), while the correlation of WC with intra-abdominal fat area was stronger (r = 0.52) [14] . In our Japanese-American population, the correlation of VAI with intra-abdominal fat area was 0.29 for men and 0.38 for women, while the correlation of WC with intra-abdominal fat area was higher (0.69 for men, 0.75 for women), but the measure with the highest correlation with intra-abdominal fat area was VAE (0.78 for men and 0.85 for women).In a study of 180 premenopausal Korean women with polycystic ovarian syndrome, VAI displayed an AUROC of 0.88 in the prediction of CT-measured VAT >100 cm&sup2; [10] , which was considerably greater than the AUROC we observed in either of our cohorts. Results from the secondary analyses we performed suggest a potential mechanism for some of the underlying differences in the relative performance of these measures. In 1991, Despr s et al. used all-possible-regressions selection methods to arrive at an adiposity surrogate to predict intra-abdominal fat area in men that incorporated age and WC [21] . Increasing age is a recognized determinant of visceral adiposity [25] , potentially due to metabolic changes in visceral fat depots [26] . A prospective study of this Japanese-American population demonstrated that weight gain at older ages was more likely to accumulate fat in the visceral depot [27] . Age may therefore be a key feature associated with visceral fat deposition, which merits consideration for inclusion in surrogate measures of this depot. The other adiposity surrogates that we tested did not directly account for age. This study has several strengths. Chief among them, we used two populations of Japanese participants with well-characterized measurements of anthropometric and metabolic features. Secondly, the novel index that we developed was externally validated and found to perform similarly with regard to capturing the visceral adipose depot in both populations. Lastly, the small number of readily available covariates that were used in the model allows generalizability to external datasets and routine clinical use. There are also some limitations. First, measurements of intra-abdominal fat area were derived from single CT-scan slices; however, although the strength of the relationship varies somewhat depending on the location of the CT slice [28] , a high correlation has been demonstrated between a single CT slice and direct measurement of visceral fat volume, which limits the potential for bias [29] . Second, we did not assess the contribution of menopausal status on intra-abdominal fat area in women. Conclusion In conclusion, accurate estimation of intra-abdominal fat area is of special importance in individuals of Asian ancestry, as they are more prone than Caucasiansto visceral fat deposition [30,31] ; can have a higher body fat percentage for a given BMI [32] ; and are at greater risk to develop diabetes at any given body-mass index [33] . When used as designed [34] , VAI is associated with metabolic dysfunction in a variety of populations and disease states. It is not, however, consistently related to amount of visceral adipose tissue. Alternative measures such as VAE may provide better estimates of intra-abdominal fat area in individuals of Japanese ancestry and potentially in other Asian and perhaps non-Asian populations as well. Further testing in other populations would demonstrate its potential value to achieve this aim. Funding: NIH K08DK103945, DK-031170, HL-049293, DK- 002654, DK-017047, DK-035816 and RR-000037; JSPS KAKENHI Grant Numbers 17390177, 20390187, and 23390177. Contribution Statement: P.L.W. researched the data and wrote the manuscript. E.J.B., T.H., K.K.S., and D.L.L. collected and researched the data, contributed to the discussion, and edited the manuscript. S.U. reviewed and edited the manuscript. Y.H. collected the data, and reviewed the manuscript. W.Y.F. collected and researched the data, contributed to the discussion, and edited the manuscript. E.J.B., T.H., and P.L.W. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Conflict of interest: None Supplementary Table 1: Model R 2 from linear regression models predicting continuous intra-abdominal fat area, and AUROC and 95% CI for prediction of visceral adiposity category at median and 75 th percentile cutpoints by logistic regression models (alternate cutpoints). Japanese American Community Diabetes Study Ohtori Study Men Women Men Women n = 344 n = 278 n = 1033 n = 612 R&sup2; AUROC (95% CI) p-value R&sup2; AUROC (95% CI) p-value R&sup2; AUROC (95% CI) p-value R&sup2; AUROC (95% CI) p-value Prediction of visceral adiposity, linear model VAE 0.59 0.72 0.59 0.64 VAI 0.08 0.14 0.05 0.23 WC 0.48 0.56 0.55 0.56 Prediction of visceral adiposity category at median cutpoints VAE 0.88 (0.84 - 0.91) - 0.94 (0.92 - 0.97) - 0.88 (0.86 - 0.90) - 0.91 (0.89 - 0.94) - VAI 0.74 (0.69 - 0.79) < 0.001 0.79 (0.73 - 0.84) < 0.001 0.72 (0.68 - 0.75) < 0.001 0.79 (0.75 - 0.83) < 0.001 WC 0.81 (0.77 - 0.86) < 0.001 0.88 (0.84 - 0.92) < 0.001 0.86 (0.84 - 0.88) 0.026 0.86 (0.83-0.89) < 0.001 Prediction of visceral adiposity category at 75 th percentile cutpoints VAE 0.88 (0.84 - 0.92) - 0.93 (0.90 - 0.96) - 0.87 (0.85-0.89) - 0.93 (0.91 - 0.95) - VAI 0.73 (0.67 - 0.79) < 0.001 0.78 (0.72 - 0.83) < 0.001 0.72 (0.69 - 0.76) < 0.001 0.78 (0.74 - 0.82) < 0.001 WC 0.84 (0.79 - 0.88) 0.004 0.88 (0.84 - 0.92) 0.001 0.86 (0.83 - 0.88) 0.0539 0.89 (0.86-0.92) 0.001 Abbreviations: AUROC (area under the receiver-operating characteristics curve), VAE (visceral adiposity estimate), VAI (visceral adiposity index), WC (waist circumference) *For JACDS, the median intra-abdominal fat area was 100 cm&sup2; for men and 70 cm&sup2; for women. For participants in the Ohtori Study, the median intra-abdominal fat area was 77.93 cm&sup2; for men and 35.40 cm&sup2; for women. For JACDS, the 75 th percentile of intra-abdominal fat area was 140 cm&sup2; for men and 100 cm&sup2; for women. For participants in the Ohtori Study, the 75 th percentile was 107.3 cm&sup2; for men and 63.4 cm&sup2; for women. Supplementary Table 2: Comparison of extended list of adiposity surrogates in prediction of continuous intra-abdominal fat area and visceral adiposity category (low vs. high visceral adiposity) in the Japanese-American Community Diabetes Study. Men Women n = 344 n = 278 R&sup2; AUROC(95% CI) p-value* R&sup2; AUROC(95% CI) p-value* Prediction of visceral adiposity, linear model VAE 0.59 0.72 VAI 0.08 0.14 WC 0.48 0.56 Fitted VAI 0.08 0.13 Despr s 0.50 -- Fitted Despr s 0.57 0.68 Fitted EWET 0.25 0.09 LAP 0.17 0.22 DAAT 0.46 0.58 Prediction of visceral adiposity category at median cutpoints VAE -- 0.88 (0.84 - 0.91) -- 0.94 (0.92 - 0.97) -- VAI a 0.74 (0.69 - 0.79) < 0.001 0.79 (0.73 - 0.84) < 0.001 WC 0.81 (0.77 - 0.86) < 0.001 0.88 (0.84 - 0.92) < 0.001 Fitted VAI a 0.74 (0.69 - 0.79) < 0.001 0.78 (0.72 - 0.83) < 0.001 Despr s adipose-tissue estimate b 0.85 (0.81-0.89) 0.006 - - Fitted Despr s adipose-tissue estimate c 0.86 (0.83 - 0.90) 0.029 0.93 (0.91 - 0.96) 0.225 Fitted EWET d 0.63 (0.60 - 0.67) < 0.001 0.54 (0.52 - 0.56) < 0.001 LAP index e 0.79 (0.75 - 0.84) < 0.001 0.87 (0.83 - 0.91) < 0.001 DAAT index f 0.80 (0.75 - 0.85) < 0.001 0.89 (0.85 - 0.93) < 0.001 Prediction of visceral adiposity category at 75 th percentile cutpoints VAE 0.88 (0.84 - 0.92) - 0.93 (0.90 - 0.96) - VAI a 0.73 (0.67 - 0.79) < 0.001 0.78 (0.72-0.83) < 0.001 WC 0.84 (0.79 - 0.88) 0.002 0.88 (0.84 - 0.92) < 0.001 Fitted VAI a 0.73 (0.66 - 0.79) < 0.001 0.76 (0.70 - 0.82) < 0.001 Despr s adipose-tissue estimate b 0.85 (0.80 - 0.89) 0.006 - - Fitted Despr s adipose-tissue estimate c 0.87 (0.83 - 0.91) 0.218 0.90 (0.87 - 0.94) 0.003 Fitted EWET d 0.69 (0.64 - 0.75) < 0.001 0.56 (0.52 - 0.60) < 0.001 LAP index e 0.80 (0.75 - 0.85) < 0.001 0.86 (0.81 - 0.90) < 0.001 DAAT index f 0.82 (0.77 - 0.87) < 0.001 0.88 (0.84 - 0.92) < 0.001 Abbreviations: AUROC (area under the receiver-operating characteristics curve), VAE (visceral adiposity estimate), VAI (visceral adiposity index), WC (waist circumference) * p-value compares AUROC to VAE AUROC a calculated de novo by regressing WC on BMI, setting the ratio of BMI:WC equal to 1, then multiplying by triglyceride level in mmol/l divided by median triglyceride level and median HDL level in mmol/l divided by the HDL level, following the method of Amato b calculated using coefficients derived by Despr s et al. (1991) in 110 men: IAFA = -225.39 + 2.125 x age (years) + 2.843 x WC(cm), available for men only c calculated de novo using fitted values from JACDS, in sex-specific linear regression models of IAFA on age and WC d Combined enlarged waist with elevated triacylgylcerols (EWET) = yes for men if WC > 95 cm and triglyceride > 1.45 mmol/L / = yes for women if WC > 88 cm and triglyceride > 1/45 mmol/L, from Kahn et al. (2003) [22] e Lipid accumulation product (LAP) index, men [(WC (cm)-65) x TG (mmol)], women [(WC (cm) -58) x TG (mmol)], in the manner of Wakabayashi et al. (2014) [23] f Deep-abdominal adiposity tissue (DAAT), men = 382.9 + [1.09 x weight -(kg)] + [6.04 x WC-(cm)] + (-2.29 x BMI), women = -278 + [-0.86 x weight -(kg)] + [5.19 x WC-(cm)] , in the manner of Brundavani et al. (2006) [24] References 1. Hayashi, T., Boyko, E.J., McNeely, M.J., et al. Visceral adiposity, not abdominal subcutaneous fat area, is associated with an increase in future insulin resistance in Japanese Americans. (2008) Diabetes 57(5): 1269-1275. 2. Fujimoto, W.Y., Bergstrom, R.W., Boyko, E.J., et al. Visceral adiposity and incident coronary heart disease in Japanese-American men. The 10-year follow-up results of the Seattle Japanese-American Community Diabetes Study. (1999) Diabetes Care 22(11): 1808-1812. 3. Hayashi, T., Boyko, E.J., Leonetti, D.L., et al. Visceral adiposity is an independent predictor of incident hypertension in Japanese Americans. (2004) Ann Intern Med140(12): 992-1000. 4. Hayashi, T., Boyko, E.J., Leonetti, D.L., et al. Visceral adiposity and the risk of impaired glucose tolerance: a prospective study among Japanese Americans. (2003) Diabetes care 26(3): 650-655. 5. Boyko, E.J., Fujimoto, W.Y., Leonetti, D.L., et al. Visceral adiposity and risk of type 2 diabetes: a prospective study among Japanese Americans. (2000) Diabetes Care 23(4): 465-471. 6. Tong, J., Boyko, E.J., Utzschneider, K.M., et al. Intra-abdominal fat accumulation predicts the development of the metabolic syndrome in non-diabetic Japanese-Americans. (2007) Diabetologia 50(6): 1156-1160. 7. Nazare, J.A., Smith, J., Borel, A.L., et al. Usefulness of measuring both body mass index and waist circumference for the estimation of visceral adiposity and related cardiometabolic risk profile (from the INSPIRE ME IAA study). (2015) Am J Cardiol 115(3): 307-315. 8. Amato, M.C., Giordano, C., Galia, M., et al. Visceral Adiposity Index: a reliable indicator of visceral fat function associated with cardiometabolic risk. (2010) Diabetes Care 33(4): 920-922. 9. Musso, G., Cassader, M., Gambino, R. Diagnostic accuracy of adipose insulin resistance index and visceral adiposity index for progressive liver histology and cardiovascular risk in nonalcoholic fatty liver disease. (2012) Hepatology 56(2): 788-789. 10. Oh, J.Y., Sung, Y.A., Lee, H.J. The visceral adiposity index as a predictor of insulin resistance in young women with polycystic ovary syndrome. (2013) Obesity (Silver Spring) 21(8): 1690-1694. 11. Bozorgmanesh, M., Hadaegh, F., Azizi F. Predictive performance of the visceral adiposity index for a visceral adiposity-related risk: type 2 diabetes. (2011) Lipids Health Dis10: 88. 12. Chen, C., Xu,Y., Guo, Z.R., et al. The application of visceral adiposity index in identifying type 2 diabetes risks based on a prospective cohort in China. (2014) Lipids Health Dis 13: 108. 13. Du, T., Sun, X., Huo, R., et al. Visceral adiposity index, hypertriglyceridemic waist and risk of diabetes: the China Health and Nutrition Survey 2009. (2014) Int J Obes (Lond) 38(6): 840-847. 14. Vongsuvanh, R., George, J., McLeod, D., et al. Visceral adiposity index is not a predictor of liver histology in patients with non-alcoholic fatty liver disease. (2012) J Hepatol 57(2): 392-398. 15. Elisha, B., Messier, V., Karelis, A., et al. The Visceral Adiposity Index: Relationship with cardiometabolic risk factors in obese and overweight postmenopausal women - A MONET group study. (2013) Appl Physiol Nutr Metab 38(8): 892-899. 16. Fujimoto, W.Y., Leonetti, D.L., Kinyoun, J.L., et al. Prevalence of diabetes mellitus and impaired glucose tolerance among second-generation Japanese-American men. (1987) Diabetes 36(6): 721-729. 17. Koh, H., Hayashi, T., Sato, K.K., et al. Visceral adiposity, not abdominal subcutaneous fat area, is associated with high blood pressure in Japanese men: the Ohtori study. (2011) Hypertens Res 34(5): 565-572. 18. Shuman, W.P., Morris, L.L., Leonetti, D.L., et al. Abnormal body fat distribution detected by computed tomography in diabetic men. (1986) Invest Radiol 21(6): 483-487. 19. Moran, J.L., Solomon, P.J. Statistics in review Part I: graphics, data summary and linear models. (2007) Crit Care Resusc 9(1): 81-90. 20. Pepe, M.S. The statistical evaluation of medical tests for classification and prediction. (2003) Oxford University Press: Oxford ; New York, 2003. 21. Despres, J.P., Prud'homme, D., Pouliot, M.C., et al. Estimation of deep abdominal adipose-tissue accumulation from simple anthropometric measurements in men. (1991) Am J Clin Nutr 54(3): 471-477. 22. Kahn, H.S., Valdez, R. Metabolic risks identified by the combination of enlarged waist and elevated triacylglycerol concentration. (2003) Am J Clin Nutr 78(5): 928-934. 23. Wakabayashi, I., Daimon, T. A strong association between lipid accumulation product and diabetes mellitus in japanese women and men. (2014) J Atheroscler Thromb 21(3): 282-288. 24. Brundavani, V., Murthy, S.R., Kurpad, A.V. Estimation of deep-abdominal-adipose-tissue (DAAT) accumulation from simple anthropometric measurements in Indian men and women. (2006) Eur J Clin Nutr 60(5): 658-666. 25. Bray, G.A., Bouchard, C. Handbook of obesity: epidemiology, etiology, and physiopathology, 3rd edn (2014) CRC Press: Boca Raton. 26. Park, S.E., Park, C.Y., Choi, J.M., et al. Depot-Specific Changes in Fat Metabolism with Aging in a Type 2 Diabetic Animal Model. (2016) PloS One 11(2): e0148141. 27. Lee, C.G., Fujimoto, W.Y., Brunzell, J.D., et al. Intra-abdominal fat accumulation is greatest at younger ages in Japanese-American adults. (2010) Diabetes Res Clin Pract 89(1): 58-64. 28. Shen, W., Punyanitya, M., Wang, Z., et al. Visceral adipose tissue: relations between single-slice areas and total volume. (2004) Am J Clin Nutr 80(2): 271-278. 29. Kelley, D.E., Thaete, F.L., Troost, F., et al. Subdivisions of subcutaneous abdominal adipose tissue and insulin resistance. (2000) Am J Physiol Endocrinol Metab 278(5): E941-E948. 30. Park, Y.W., Allison, D.B., Heymsfield, S.B., et al. Larger amounts of visceral adipose tissue in Asian Americans. (2001) Obes Res 9(7): 381-387. 31. Lim, U., Ernst, T., Buchthal, S.D., et al. Asian women have greater abdominal and visceral adiposity than Caucasian women with similar body mass index. (2011) Nutr Diabetes 1(5): e6. 32. Deurenberg, P., Deurenberg-Yap, M., Guricci, S. Asians are different from Caucasians and from each other in their body mass index/body fat per cent relationship. (2002) Obes Rev 3(3): 141-146. 33. McNeely, M.J., Boyko, E.J. Type 2 diabetes prevalence in Asian Americans: results of a national health survey. (2004) Diabetes Care 27(1): 66-69. 34. Amato, M.C., Giordano, C. Clinical indications and proper use of Visceral Adiposity Index. (2013) Nutr, Metab Cardiovascular Dis 23(8): e31-e32.