Diabetes is one of the fastest growing health challenges of the twenty-first century with the number of adults living with diabetes more than tripling over the past 20 years(Reference Fox, Massaro and Hoffmann1). In 2019, the International Diabetes Federation estimated that 463 million adults had diabetes globally. In China, the number of people with diabetes reached 116 million in 2019 and is expected to reach 147·2 million in 2045(2). Type 2 diabetes mellitus (T2DM), which is characterised by insulin resistance, accounts for 90 % of all diabetes(Reference Festa, Williams and D’Agostino3) and is becoming one of the world’s leading disease burdens, especially in China(Reference Vos, Flaxman and Naghavi4,5) . 55·9 % of people with diabetes in China are still undiagnosed, however(2); therefore, early and accurate identification of T2DM-related risk factors is urgently needed to effectively reduce the incidence and disease burden of T2DM into the future.
The Metabolic Score for Visceral Fat (METS-VF) is a novel estimator which combines measures of fasting plasma glucose (FPG), TAG, BMI, HDL-cholesterol, waist to height ratio (WHtR), age and sex to estimate visceral adipose tissue (VAT) and predict the incidence of cardiometabolic complications(Reference Bello-Chavolla, Antonio-Villa and Vargas-Vázquez6). In a Mexican urban population, METS-VF was validated as a better performing predictor of T2DM incidence compared with other surrogate VAT indexes, including visceral adiposity index (VAI) and lipid accumulation product(Reference Bello-Chavolla, Antonio-Villa and Vargas-Vázquez6). One study to date, in southern Indian individuals with morbid obesity(Reference Kapoor, Jiwanmall and Nandyal7), has validated the utility of METS-VF as a surrogate measure of visceral adiposity, while another retrospective cohort study from China indicated that METS-VF could be a useful tool for the hierarchical prevention and management of hyperuricaemia among non-obese women(Reference Liu, Chen and Xu8). So far, however, no published studies have explored the relationship between METS-VF and T2DM risk in the Chinese population.
This study therefore aimed to evaluate the association of the METS-VF with incidence of T2DM and to compare the predictive value of the METS-VF for T2DM incidence with that of other obesity indices (VAI, a body shape index (ABSI), WHtR, waist circumference (WC) and BMI) in the Chinese population.
Subjects and methods
Participants and study setting
This was a prospective cohort study with 20 194 participants aged ≥ 18 years, who were recruited during July to August 2007 and July to August 2008 from a rural area in Henan Province in China(Reference Zhang, Zhao and Sun9). All participants were investigated using an interview questionnaire, anthropometric measurements and laboratory measurements. All participants were free of severe psychological disorders, physical disabilities, Alzheimer disease, dementia, tuberculosis, AIDS or other infectious diseases at the time of enrolment. Follow-up examination of 17 265 participants was performed between July and August 2013 and from July to October 2014 (response rate was 85·5 %). All individuals signed an informed consent form, approved by the Ethics Committee of Zhengzhou University.
During follow-up, the same interview questionnaire was administered again and the same measurements taken as for the baseline examination, including questionnaire interview, anthropometric measurements and laboratory measurements.
In the current analysis, we have excluded participants with diabetes (T2DM, T1DM, gestational diabetes and other special types of diabetes) at baseline (n 1512), those whose diabetes status was unknown at baseline or follow-up (n 3477) and those with missing data for baseline age, weight, height, WC, FPG, TAG or HDL-cholesterol (n 39). Finally, a total of 12 237 individuals were included in the study.
Data collection
Interview questionnaire
Well-trained research staff collected demographic information (age, sex and education level), lifestyle data (smoking, alcohol drinking and physical activity) and medical history (family history of disease and personal history of disease) during face-to-face interviews, using a standard questionnaire. The definition of smoker, alcohol drinking and physical activity level were in accordance with the international standards, which has been described in detail in the previous studies(Reference Han, Liu and Sun10,Reference Craig, Marshall and Sjöström11) .
Anthropometric measurements
All participants were asked to wear light clothing and no shoes for anthropometric measurement. Height was measured to the nearest 0·1 cm. Weight was measured to the nearest 0·5 kg. WC was measured at the midpoint between the lowest rib and the iliac crest to the nearest 0·1 cm. All measurements were taken twice and recorded as the average of the two. Before blood pressure measurement, participants were instructed to refrain from smoking and from drinking alcohol, coffee and tea. Blood pressure was measured three times using an electronic sphygmomanometer device (HEM-770AFuzzy; Omron) on the unclothed right upper arm, at intervals of 30 s, with the three measurements recorded, according to the American Heart Association’s standardised protocol(Reference Perloff, Grim and Flack12).
Laboratory measurements
Fasting blood samples were obtained after an overnight fast of at least 8 h. Levels of FPG, TAG, total cholesterol and HDL-cholesterol were measured using a HITACHI automatic clinical analyser (Model 7060). The Friedewald formula was used to calculate serum concentrations of LDL-cholesterol(Reference Bairaktari, Hatzidimou and Tzallas13). METS-VF(Reference Bello-Chavolla, Antonio-Villa and Vargas-Vázquez6), VAI(Reference Amato, Giordano and Galia14), ABSI(Reference Krakauer and Krakauer15), WHtR and BMI were calculated as follows:
$${\rm{METS}} - {\rm{VF}}\left( {{\rm{men}}} \right) = {\rm{4}}.{\rm{466}} + 0.0{\rm{11}}$$
$$*({\rm{( Ln}}\left( {{\rm{(Ln}}\left( {\left( {{\rm{2}}*{\rm{FPG}}} \right) + {\rm{TG}}} \right)*{\rm{BMI}}} \right)/\left( {{\rm{Ln}}\left( {{\rm{HDL}} - {\rm{C}}} \right)} \right)))**{\rm{3}})$$
$$ + {\rm{3}}.{\rm{239}}*\left( {\left( {{\rm{Ln}}\left( {{\rm{WHtR}}} \right)} \right)**{\rm{3}}} \right) + 0.{\rm{319}}*{\rm{1}} + 0.{\rm{594}}*\left( {{\rm{Ln}}\left( {{\rm{age}}} \right)} \right)$$
$${\rm{METS}} - {\rm{VF}}\left( {{\rm{women}}} \right) = {\rm{4}}.{\rm{466}} + 0.0{\rm{11}}$$
$$*{\rm{((Ln(}}\left( {{\rm{Ln}}\left( {\left( {{\rm{2}}*{\rm{FPG}}} \right) + {\rm{TG}}} \right)*{\rm{BMI}}} \right)/\left( {{\rm{Ln}}\left( {{\rm{HDL}} - {\rm{C}}} \right)} \right){\rm{)}}**{\rm{3)}}$$
$$ + {\rm{3}}.{\rm{239}}*\left( {\left( {{\rm{Ln}}\left( {{\rm{WHtR}}} \right)} \right)**{\rm{3}}} \right) + 0.{\rm{319}}*0 + 0.{\rm{594}}$$
$$*\left( {{\rm{Ln}}\left( {{\rm{age}}} \right)} \right)$$
$${\rm{VAI}}\left( {{\rm{men}}} \right) = \left( {{\rm{WC}}/\left( {{\rm{39}}.{\rm{68}} + \left( {{\rm{1}}.{\rm{88}}*{\rm{BMI}}} \right)} \right)} \right)*\left( {{\rm{TG}}/{\rm{1}}.0{\rm{3}}} \right)*\left( {{\rm{1}}.{\rm{31}}/{\rm{HDL}} - {\rm{C}}} \right)$$
$${\rm{VAI}}\left( {{\rm{women}}} \right) = \left( {{\rm{WC}}/\left( {{\rm{36}}.{\rm{58}} + \left( {{\rm{1}}.{\rm{88}}*{\rm{BMI}}} \right)} \right)} \right)*\left( {{\rm{TG}}/{\rm{0}}.8{\rm{1}}} \right)*\left( {{\rm{1}}.{\rm{52}}/{\rm{HDL}} - {\rm{C}}} \right)$$
$${\rm{ABSI}} = {\rm{WC}}/\left( {\left( {{\rm{BMI}}**\left( {{\rm{2}}/{\rm{3}}} \right)} \right)*\left( {{\rm{height}}**\left( {{\rm{1}}/{\rm{2}}} \right)} \right)} \right)$$
$${\rm{WHtR}} = {\rm{WC}}\left( {{\rm{cm}}} \right)/{\rm{height}}\left( {{\rm{cm}}} \right)$$
$${\rm{BMI}} = {\rm{weight}}\left( {{\rm{kg}}} \right)/{\rm{heigh}}{{\rm{t}}^{\rm{2}}}\left( {{{\rm{m}}^{\rm{2}}}} \right)$$
Type 2 diabetes definition
T2DM was considered FPG ≥ 7·0 mmol/l and/or use of insulin or oral hypoglycaemic medication and/or a self-reported history of T2DM after excluding gestational diabetes, T1DM and other special types of diabetes(16).
Statistical analyses
Baseline demographic characteristics were described with continuous data summarised as median (interquartile range) for the skewed distribution, and for categorical variables as frequency (%). χ 2 or Kruskal–Wallis tests were used to compare differences among groups. The cox proportional hazards regression was used to estimate the risk of T2DM by calculating the hazard ratios and 95 % CI, with the lowest quartile of METS-VF as the reference. We also estimated the risk of T2DM with a per 1-sd increase in METS-VF. Model 1 was unadjusted; model 2 was adjusted for age and sex; model 3 was adjusted for model 2 plus education level, marital status, physical activity, tobacco, alcohol and family history of T2DM; and model 4 was adjusted for model 3 plus systolic blood pressure, diastolic blood pressure, total cholesterol and LDL-cholesterol at baseline. A sensitivity analysis was also performed to verify the robustness of our results by excluding participants with cancer, kidney disease, stroke, myocardial infarction or heart failure at baseline. We also performed subgroup analyses stratified by sex (men and women), age (< 45 and ≥ 45 years) and BMI (BMI < 24 and ≥ 24 kg/m2) at baseline after adjusting for the potential confounding factors in model 4. We used restricted cubic splines with four knots at the 5th, 25th, 75th and 95th centiles to flexibly model the association between METS-VF and T2DM risk, with the knot at the 25th percentile of the distribution as the reference.
Finally, the area under the receiver operating characteristic curve (AUC) was used to test the ability of baseline METS-VF, VAI, ABSI, WHtR, WC and BMI to predicting T2DM incidence at follow-up. The Z statistic was used to test differences between the AUC.
The receiver operating characteristics were calculated using Medcalc V9.3, and restricted cubic splines were performed with R.3.6.3 (R Foundation), while other analyses involved using SAS V9.4 for Windows (SAS Inst.). All statistical analyses were considered statistically significant, with two-sided P < 0·05.
Results
Demographic characteristics of the study population
Table 1 shows the baseline characteristics of the study participants with and without T2DM during follow-up. A total of 12 237 participants (4646 or 37·97 % men) were included in this study. Participants who developed T2DM were older, with a median age (interquartile range) of 53 (44–60) for T2DM and 50 (41–59) for non-T2DM, respectively. Values for systolic blood pressure, diastolic blood pressure, FPG, total cholesterol, TAG, LDL-cholesterol, METS-VF, VAI, ABSI, WHtR, WC and BMI were higher, while HDL-cholesterol levels were lower for those with T2DM than for those without (P < 0·001).
Baseline characteristics of the included participants
(Numbers and percentages; median and interquartile ranges)
SBP, systolic blood pressure; DBP, diastolic blood pressure; FPG, fasting blood glucose; TC, total cholesterol; METS-VF, Metabolic Score for Visceral Fat; VAI, visceral adiposity index; ABSI, a body shape index; WHtR, waist-to-height ratio; WC, waist circumference.
Association of baseline Metabolic Score for Visceral Fat and type 2 diabetes mellitus risk
During a median follow-up of 6·01 (5·09–6·06) years, T2DM developed in 837 cases among the 12 237 participants. The incidence densities of T2DM were 3·55, 8·07, 12·42 and 22·63 per 1000 person-years for quartiles 1, 2, 3 and 4 of METS-VF, respectively (Table 2). In unadjusted model 1, with METS-VF quartile 1 as the reference, the hazard ratios for T2DM were 2·42 (95 % CI 1·81, 3·24), 3·96 (95 % CI 3·01, 5·22) and 7·74 (95 % CI 5·95, 10·07) for quartiles 2, 3 and 4, respectively. After adjusting for potential confounding factors, the positive association between METS-VF and T2DM risk persisted. In model 4, the hazard ratios for T2DM with quartiles 2, 3 and 4 of METS-VF were 2·18 (95 % CI 1·56, 3·07), 3·16 (95 % CI 2·27, 4·40) and 5·97 (95 % CI 4·28, 8·32), respectively, with quartile 1 as reference. T2DM risk significantly increased with higher quartiles of METS-VF (P trend < 0·001).
Cox proportional hazard regression analysis of the association of baseline metabolic score for visceral fat and risk of type 2 diabetes mellitus
(Hazard ratios and 95 % confidence intervals)
* Per 1000 person-years.
Model 1: Unadjusted.
Model 2: Adjusted for age and sex at baseline.
Model 3: Adjusted for age, sex, physical activity, tobacco, alcohol, education level, marital status and family history of type 2 diabetes mellitus at baseline.
Model 4: Adjusted for age, sex, physical activity, tobacco, alcohol, education level, marital status, family history of type 2 diabetes mellitus at baseline as well as systolic blood pressure, diastolic blood pressure, total cholesterol, LDL-cholesterol at baseline.
Sensitivity analysis: Adjusted for model 4 and further excluded participants with cancer, kidney disease, stroke, myocardial infarction or heart failure at baseline.
Risk of T2DM was increased with a per 1-sd increase in METS-VF for all participants. The adjusted hazard ratio was 2·15 (95 % CI 1·90, 2·44) in model 4. Restricted cubic splines indicated a significant nonlinear dose–response association between METS-VF and T2DM risk for all participants (P nonlinearity = 0·0347, Fig. 1)
Dose–response association between Metabolic Score for Visceral Fat (METS-VF) and risk of type 2 diabetes mellitus (P nonlinearity = 0·0347).
Sensitivity analysis and subgroup analyses for the association of Metabolic Score for Visceral Fat and type 2 diabetes mellitus risk
In sensitivity analysis, the positive association between METS-VF and T2DM risk was essentially robust after excluding participants with cancers, kidney disease, stroke, myocardial infarction or heart failure at baseline, respectively (Table 2). In the subgroup analyses stratified by sex, age and BMI, after adjusting potential confounding factors in model 4 and with quartile 1 as reference, the risk of T2DM in the highest quartile still remained significant (Fig. 2).
Subgroup analyses of the association between metabolic score for visceral fat and risk of type 2 diabetes mellitus. HR, hazard ratio; Q, quartile.
Comparison of the association of Metabolic Score for Visceral Fat, visceral adiposity index, a body shape index, waist to height ratio, waist circumference and BMI with type 2 diabetes mellitus risk
Table 3 shows AUC (95 % CI), the optimal cut-offs and corresponding sensitivity, specificity, positive likelihood ratio (+LR), negative likelihood ratio (–LR), positive predictive value (+PV), negative predictive value (–PV) and Youden index for each index predicting T2DM risk. The receiver operating characteristic curve analyses for METS-VF, VAI, ABSI, WHtR, WC and BMI for predicting T2DM incidence are shown in online Supplementary Fig. S1. The AUC for METS-VF, VAI, ABSI, WHtR, WC and BMI were 0·690 (95 % CI 0·682, 0·698), 0·657 (95 % CI 0·649, 0·666), 0·621 (95 % CI 0·613, 0·630), 0·681 (95 % CI 0·672, 0·689), 0·680 (95 % CI 0·672, 0·689) and 0·661 (95 % CI 0·652, 0·669), respectively. METS-VF had the largest AUC for predicting T2DM incidence, significantly different from VAI, ABSI and BMI (P < 0·05) but not WHtR or WC (P = 0·058). Compared with VAI (0·25), ABSI (0·19), WHtR (0·28), WC (0·27) and BMI (0·25), the Youden index for METS-VF was the highest for all participants (0·29).
Comparison of areas under the receiver operating characteristic curves for METS-VF, VAI, ABSI, WC and BMI with type 2 diabetes mellitus risk
(95 % confidence intervals)
METS-VF, Metabolic Score for Visceral Fat; METS-IR, Metabolic Score for IR; VAI, visceral adiposity index; ABSI, a body shape index; WHtR, waist-to-height ratio; WC, waist circumference; +LR, positive likelihood ratio; –LR, negative likelihood ratio; +PV, positive predictive value; –PV, negative predictive value.
Discussion
In this large prospective cohort study, we explored the association between METS-VF and the risk of T2DM in a rural Chinese population. Our results showed a positive association of METS-VF with T2DM and a significant increase in T2DM incidence across quartiles of METS-VF after adjusting for potential risk factors. Similar results were observed in the sensitivity analysis. This positive association was also found in subgroup analyses by sex, age and BMI. Additionally, we found a nonlinear association between METS-VF and T2DM. Moreover, among the five indices (METS-VF, VAI, ABSI, WHtR, WC and BMI), METS-VF showed the largest AUC and the highest Youden index in predicting the risk of T2DM.
An increase in fat mass is considered to be an important risk factor for the incidence of T2DM worldwide(Reference Stefan17–Reference Chen, Hou and Hu20), especially VAT mass which has been shown to be more harmful than fat stored elsewhere in the body(Reference Fox, Massaro and Hoffmann1,Reference Vega, Adams-Huet and Peshock21,Reference Yuan and Larsson22) . Previous studies have also shown that the risk factors associated with diabetes were more related to visceral fat than to the accumulation of systemic fat, especially in East Asian populations who are generally less obese than people in Western countries(Reference Yoon, Lee and Kim23,Reference Sjöström24) . These reports suggested the importance of measuring visceral fat accumulation. Clinical standards for assessing VAT include MRI, computerised tomography and dual X-ray absorptiometry, but they are expensive, need to be performed and interpreted by a specialist, and are often limited by equipment and technical difficulties(Reference Bello-Chavolla, Antonio-Villa and Vargas-Vázquez6,Reference Shuster, Patlas and Pinthus25) . A cheap and convenient indicator to measure the VAT is therefore needed.
Routinely applicable anthropometrical indicators of VAT content include WC, BMI and WHtR, but these indicators are of limited value because subcutaneous adipose tissue and VAT cannot be clearly distinguished(Reference Ashwell, Cole and Dixon26–Reference Chen, Xu and Guo31). VAI, based on BMI, WC, TAG and HDL-cholesterol, was established to estimate VAT accumulation predicting cardiometabolic risk in Italians(Reference Amato, Giordano and Galia14), but was not superior to simple obesity indices (BMI and WC) in predicting incidence of T2DM in Chinese people(Reference Deurenberg, Deurenberg-Yap and Guricci32–Reference Zhang, Zheng and Li35). Our study adopted METS-VF, as a novel estimator of VAT, including the main predictors of VAT (insulin resistance, WHtR, age and sex(Reference Amato, Giordano and Galia14,Reference Brundavani, Murthy and Kurpad36,Reference Lee, Carr and Murdoch37) ), to predict T2DM incidence. Our results showed that METS-VF has the strongest association with the risk of T2DM and had the largest AUC and the highest Youden index in predicting T2DM incidence compared with other indices (VAI, ABSI, WHtR, WC and BMI).
Our analysis showed that a per 1-sd gain in METS-VF was positively associated with the risk of T2DM for all participants after adjusting for potential confounders. Consistent with our findings, Neeland et al. found visceral fat mass was independently associated with T2DM among obese individuals (OR 2·40; 95 % CI 1·60, 3·70)(Reference Neeland, Turer and Ayers38). Similarly, a prospective study conducted among Japanese Americans found that intra-abdominal fat was a significant predictor of T2DM risk (OR 1·60; 95 % CI 1·10, 2·30)(Reference Boyko, Fujimoto and Leonetti39). In addition, previous studies have suggested that visceral fat was associated with T2DM independent of BMI(Reference Wei, Gaskill and Haffner40,Reference Ohlson, Larsson and Svärdsudd41) . Our subgroup analyses showed that participants in the highest quartile of METS-VF had a significantly increased risk of T2DM both with normal weight (BMI < 24 kg/m2) and if overweight/obese (BMI ≥ 24 kg/m2), compared with the lowest quartile. Consistent with our findings, a cross-sectional study from Shanghai in China that included 4126 individuals found that visceral fat significantly increased the risk of diabetes in the normal weight and overweight/obese groups(Reference Lv, Zhou and Sun42). In addition, our results showed that the positive associations occurred in both sexes, though Lv et al. found visceral fat was independently associated with T2DM only among women(Reference Lv, Zhou and Sun42). This difference may be due to the cross-sectional study design and the different target population. Further research in other populations may be needed to test the stability of METS-VF in predicting the incidence of T2DM.
The exact mechanism underlying the association of VAT and T2DM incidence is unknown, yet. Several possible mechanisms have been proposed, however. One is that VAT is an active endocrine organ in which excess visceral fat deposition causes disrupted endocrine function and dysregulation of proinflammatory factors, both of which may contribute to insulin resistance and the development of T2DM(Reference Bays43–Reference Karlsson, Rask-Andersen and Pan45). Another mechanism suggests that excess VAT may result in increased risk for T2DM through overstimulation of hepatic gluconeogenesis by chronic delivery of glycerol arising from mesenteric TAG turnover directly into the portal circulation and the liver(Reference Neeland, Hughes and Ayers46).
The strengths of this present study include longitudinal follow-up in a prospective cohort, large sample size and standardised questionnaire and laboratory procedures. We also conducted the sensitivity analysis and subgroup analyses and examined the dose–response association between baseline METS-VF and risk of T2DM. Moreover, the study first explored the association of baseline METS-VF and risk of T2DM, testing the predictive performance of T2DM among various indices (METS-VF, VAI, ABSI, WHtR, WC and BMI) in Chinese people. The AUC further confirmed the robust ability of baseline METS-VF to predict T2DM incidence at follow-up. Several limitations should be noted, however. First, we did not examine 2-h postprandial glucose, which may have led to underestimation of T2DM. Second, all the participants were rural Chinese people in this study, such that generalisation of our findings may be limited. It is yet to be determined whether METS-VF could predict the risk of T2DM in populations other than Chinese. We did not compare the results of METS-VF with actual physical measurements of visceral adiposity using MRI, computerised tomography or dual X-ray absorptiometry scans because we did not collect the relevant data. Finally, although we did adjust for known covariates, there is potential for confounding factors to affect our results given the observational design.
Conclusions
Our results show that a higher baseline METS-VF is positively associated with increased risk of T2DM in Chinese adults regardless of sex, age or BMI. Compared with other indices, METS-VF is a more effective and convenient surrogate marker for VAT measurement; it could be used in identifying the risk of T2DM in large-scale epidemiological studies.
Acknowledgements
We acknowledge all the authors and Prof. Dongsheng Hu for methodological advice and help to modify this study.
Y.F. and D.H. designed and conducted the research; F.Y. analyzed the data and wrote the paper; X.Y., Y.L., Y.W., M.H., R.Q., S.H.,X.W., Y.Z., J.Z., H.H., L.Y., T.L., D.L., F.H., M.Z., Y.Z., X.L., J.L., L.S., D.H., and Y.Z. provided constructive suggestions; and Y.F. had primary responsibility for the final content.
This study was supported by the National Natural Science Foundation of China (grant nos. 81373074, 81402752, 81673260 and 82073646); the Natural Science Foundation of Guangdong Province (grant no. 2019A1515011183) and the Natural Science Foundation of Shenzhen of China (grant no. JCYJ20190808145805515).
The authors declare that they have no conflict of interest.
Supplementary material
For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114521004116
