Skip Navigation
Skip to contents

Diabetes Metab J : Diabetes & Metabolism Journal

Search
OPEN ACCESS

Articles

Page Path
HOME > Diabetes Metab J > Ahead-of print > Article
Original Article
Metabolic Risk/Epidemiology Do Time-Dependent Repeated Measures of Anthropometric and Body Composition Indices Improve the Prediction of Incident Diabetes in the Cohort Study? Findings from a Community-Based Korean Genome and Epidemiology Study
Hye Ah Lee1orcidcorresp_icon, Hyesook Park2,3orcid, Bomi Park4orcid

DOI: https://doi.org/10.4093/dmj.2024.0357
Published online: November 15, 2024
  • 262 Views
  • 24 Download

1Clinical Trial Center, Ewha Womans University Mokdong Hospital, Seoul, Korea

2Department of Preventive Medicine, Ewha Womans University College of Medicine, Seoul, Korea

3Graduate Program in System Health Science and Engineering, Ewha Womans University, Seoul, Korea

4Department of Preventive Medicine, Chung-Ang University College of Medicine, Seoul, Korea

corresp_icon Corresponding author: Hye Ah Lee orcid Clinical Trial Center, Ewha Womans University Mokdong Hospital, 1071 Anyangcheonro, Yangcheon-gu, Seoul 07985, Korea E-mail : khyeah@naver.com
• Received: July 4, 2024   • Accepted: September 7, 2024

Copyright © 2024 Korean Diabetes Association

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

  • Background
    Cumulative evidence consistently shows that anthropometric and body composition measurements are strongly linked to the risk of incident diabetes, typically based on baseline measurements. This study aims to assess whether repeated measurements enhance the prediction of diabetes risk beyond baseline assessments alone.
  • Methods
    We utilized data from a 16-year population-based follow-up cohort within the Korean Genome and Epidemiology Study, comprising 6,030 individuals aged 40 to 69 years at baseline. We included eight indices: a body shape index (ABSI), body adiposity index (BAI), waist circumference (WC), body mass index (BMI), waist-to-hip ratio (WHR), weight-adjusted skeletal muscle index (SMI), percent body fat, and fat-to-muscle ratio. The effect of these measurements for incident diabetes was estimated using Harrell’s C-indexes and hazard ratios with 95% confidence intervals, employing time-dependent Cox proportional hazard models.
  • Results
    Over the 16-year follow-up, 939 new diabetes cases were identified (cumulative incidence, 15.6%). The median number of indicator measurements per participant was eight. The basic model, including 10 features (sex, age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline), yielded a Harrell’s C-index of 0.610. The highest C-index in repeated measurements was for WC (0.668) across the general population, weight-adjusted SMI in men, and WHR in women. However, except for ABSI and BAI, the diabetes predictive power of the other indicators was comparable. Additionally, repeated measurements of WC, BMI, and WHR in women were found to contribute to improved discrimination compared to baseline measurements, but not in men.
  • Conclusion
    Utilizing repeated measurements of general and central adiposity to predict diabetes may be helpful in predicting hidden risks, especially in women.
• Anthropometric indices are associated with diabetes but often rely on baseline data.
• Repeated measurements may uncover hidden diabetes risks, particularly in women.
• Combining abdominal and general adiposity measures enhances diabetes risk insights.
According to the 2021 Global Burden of Disease Report, approximately 460 million people worldwide were suffering from diabetes in 2019, which rose to 529 million in 2021, with type 2 diabetes mellitus (T2DM) accounting for 96.0% of these cases. Additionally, among the 16 risk factors, high body mass index (BMI) was the leading contributor to disability-adjusted life years in T2DM (52.2%), an increase of nearly 25% since 1990 [1]. Regarding the obesity epidemic, an increase in average BMI has been observed over the past decades, with the increase in BMI accelerating particularly in East and South Asia since 2000 [2].
A recent systematic review has rigorously explored the dose-response association between adiposity-related anthropometric indices and diabetes risk through cohort studies. As a result, most indicators showed a linear and monotonous association with diabetes, and the risk of diabetes increased by about 72% for every 5-unit increase in BMI and about 61% for every 10 cm increase in waist circumference (WC) [3]. In addition, the relationship with the occurrence of diabetes was also evaluated for various indicators such as visceral fat mass and subcutaneous fat mass [4], skeletal muscle index (SMI) [5], body adiposity index (BAI) [3], and fat-to-muscle ratio (FMR) [6]. These measures, reflecting various aspects of anthropometric and body composition, were found to significantly influence the progression and prognosis of diabetes [7,8].
Despite these insights, most studies traditionally focus on baseline measurements rather than exploring the potential of repeated measurements throughout the study period. A recent systematic meta-analysis based on 216 studies revealed that only a quarter of studies assessed associations with incident diabetes using repeated measures over the follow-up period [3]. This leads us to ask the question of whether baseline measurements alone are adequate to estimate future disease risk.
To address this research question, using a 16-year long-term follow-up cohort data source, we sought to estimate and compare the association between baseline versus repeated measurements of various anthropometric and body composition indices for the incident diabetes.
Data source
The study was conducted using a population-based 16-year follow-up cohort data derived from the Korean Genome and Epidemiology Study (KoGES). The community-based cohort consisted of 10,030 residents aged 40 to 69 in two communities Anseong (a rural region, n=5,018) and Ansan (an industrial region, n=5,012), starting with the baseline survey in 2001–2002, follow-up surveys are being conducted every 2 years. All participants provided written informed consent to participate in KoGES. The 10th follow-up survey was conducted in 2021–2022. More detailed information can be found in literature published elsewhere [9].
For this study, we excluded subjects with a history of cancer, cardiovascular disease (myocardial infarction, stroke, coronary artery disease, or congestive heart failure), and diabetes at baseline. We also excluded subjects with a glycosylated hemoglobin (HbA1c) level ≥6.5% or fasting blood glucose level ≥126 mg/dL. After excluding those with insufficient information on anthropometric and body composition indices, we analyzed data from 6,030 subjects (2,877 men and 3,153 women).
This study used an anonymous dataset, and the study was performed according to the guidelines of the Declaration of Helsinki and the study protocol was approved by the Ewha Womans University Hospital Institutional Review Board (IRB no. EUMC 2021-03-008).
Outcome ascertainment
The outcome of interest was the confirmation of diabetes during the follow-up period, defined by any of the following criteria: physician-diagnosed diabetes with concurrent medication use, fasting blood glucose levels of 126 mg/dL or higher, or a HbA1c of 6.5% or greater. The follow-up period commenced on the study’s start date and concluded either on the date of a physician’s diagnosis of diabetes, the first detection of fasting blood glucose ≥126 mg/dL or HbA1c ≥6.5%, or the last follow-up date, whichever came first.
Anthropometric and body composition measurements
Detailed descriptions of anthropometric and body composition measurements are provided in the Supplementary Methods. The study focused on the following eight key anthropometric and body composition indicators: a body shape index (ABSI), BAI, WC, BMI, waist-to-hip ratio (WHR), weight-adjusted SMI, percent body fat (PBF), and FMR. The calculation of anthropometric indices is described below:
ABSI=WC (cm)/[BMI0.66×height (m)0.5]
BAI=hip circumference (cm)/[(height [m]1.5) –18]
BMI=body weight (kg)/height (m)2
WHR=WC (cm)/hip circumference (cm)
SMI (%)=skeletal muscle mass (kg)/body weight (kg)×100
PBF (%)=body fat mass (kg)/body weight (kg)×100
FMR=body fat mass (kg)/skeletal muscle mass (kg)
Only the measurements up to the period before the incident diabetes were considered and used in the analysis. During the 16-year follow-up period, participants had a median of eight measurements taken for each anthropometric index.
Covariates
Based on previous studies [10-13], as features related to incident diabetes, sex, age, education levels, physical activity, alcohol intake, current smoking status, total energy intake, dietary diversity score (DDS), and log-transformed C-reactive protein levels, and unweighted genetic risk score (GRS) at baseline were considered. A detailed description can be found in the Supplementary Methods.
Statistical analysis
Descriptive statistics were estimated for the baseline characteristics of the participants, presenting mean and standard deviation (SD) or median and interquartile range based on the distribution of the numerical data. Frequencies and percentages described nominal data. Gender differences in baseline characteristics were assessed using the t-test, Mann-Whitney U test, or chi-square test as appropriate. The strength of linear relationships among anthropometric and body composition indices was determined by calculating the Pearson correlation coefficient (r) for baseline measurements and the repeated measures correlation coefficient (rrm) for time-varying measurements [14]. Intraclass correlation coefficients (ICC), obtained from linear mixed models, evaluated the reliability of repeated measurements of indices, interpreted as follows: 0.5–0.75 indicates moderate reliability, 0.75–0.9 good, and above 0.9 excellent [15].
The influence of baseline measurements of indices on the incidence of diabetes was examined using Cox proportional hazards regression models. To account for changes over time, Cox proportional regression models with time-varying covariates assessed the impact of repeated measurements of indices on diabetes risk. For these analyses, anthropometric measures were normalized so that one unit corresponded to one SD. Results were presented as hazard ratio (HR) and 95% confidence interval (CI).
Diabetes-related factors included in the analysis were sex, age, education level (did not graduate high, graduated high school, and some college or higher), physical activity (none, moderate, or heavy), alcohol intake (none, moderate, or heavy), current smoking status, total energy intake, DDS (≤5, 6–8, and ≥9), log-transformed C-reactive protein levels, and quartiles of unweighted GRS at baseline. It was named as the basic model. The statistical model for anthropometric and body composition indices was constructed by incorporating measurements of each indicator into the basic model. In terms of model fit, each indicator model was evaluated using the likelihood ratio test with reference to the basic model. The predictive probability of anthropometric and body composition indices, including various diabetes-related characteristics, for diabetes was assessed with Harrell’s C-index [16]. To control for bias due to measurement error, 500-times bootstrap was performed and results are expressed as bias-corrected Harrell’s C-index. The model with the largest Harrell’s C-index can be understood as having the best diabetes prediction accuracy [16]. We evaluated whether a model utilizing repeated measurements offered similar diabetes discrimination power compared to a model that only considered baseline measurements. Continuous net reclassification improvement (cNRI) was also calculated [17]. To determine comparability among the indicators, we also conducted pairwise comparisons (seven pairs) with other indicators, referencing the one with the highest C-index. Multiple comparisons were corrected by estimating Bonferroni-adjusted P values [16,18]. Considering gender difference in etiology [19], subgroup analysis by sex was conducted. For women, we also performed subgroup analyses by menopausal status at baseline. Additionally, to assess the necessity of follow-up measurements over several years after the baseline for predicting long-term diabetes incidence, we calculated the predictive power by substituting measurements at specific later time points with the most recent values. These results were then compared with those from a model with time-varying measurements. Sensitivity analyzes were also performed targeting individuals (n=1,444) who experienced a weight change of ≥|10%| relative to baseline measurements until the event was observed.
All statistical analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC, USA). Tests were two-tailed with a significance level set at P<0.05.
Data availability
The data described in the manuscript, code book, and analytic code will not be made available because the datasets used and/or analyzed during the current study are owned by a third party organization (The KoGES-Ansan and Ansung study; 4851-302). These data are available online with permission from the Division of Epidemiology and Health Index of the Korea Centers for Disease Control and Prevention.
Over the 16-year follow-up, 939 new diabetes cases were identified (cumulative incidence, 15.6%). Summary statistics of study subjects are detailed in Table 1. The mean BMI, a common indicator of adiposity, was 24.4 kg/m2. It was higher in women than in men (24.7 kg/m2 vs. 24.2 kg/m2). As an indicator of abdominal adiposity, the averages of ABSI, WC, and WHR were 0.1, 81.6 cm, and 0.9, respectively. As indicators reflecting body fat, BAI was 28.4, PBF was 26.7%, and FMR was 0.4. The SMI reflecting muscle mass was 69.2%. These indicators also displayed gender-specific differences.
Among the index pairs, the pairs between SMI, PBF, and FMR showed the highest correlation coefficient (rrm =–0.99), while the BMI and ABSI pair showed the lowest correlation coefficient (rrm =–0.02). Correlation coefficients between indicators were similar for baseline measurements, but the correlation coefficients for SMI and PBF with WHR were weaker than those for repeated measurements (Supplementary Fig. 1). The ICC for repeated measurement of indices were 0.904 for BMI, 0.887 for BAI, 0.850 for SMI, 0.846 for PBF, 0.836 for FMR, 0.806 for WC, 0.722 for WHR, and 0.688 for ABSI.
Tables 2 and 3 present the results of the effects of anthropometric and body composition indices on incident diabetes using baseline and repeated measurements. The basic model, which incorporated 10 features related to diabetes, achieved a Harrell’s C-index of 0.610. As repeated measurements of each indicator was added to the basic model, Harrell’s C-index improved, with WC showing the highest C-index (0.668) and ABSI the lowest (0.621). Also, this trend was maintained in the bootstrap-corrected C-index results (0.661 for WC and 0.609 for ABSI). The diabetes prediction power of WC was comparable to BMI, WHR, SMI, and PBF. The risk of incident diabetes per 1 SD for each indicator also showed the highest value in WC (HR, 1.63; 95% CI, 1.53 to 1.74) and the lowest value in ABSI (HR, 1.24; 95% CI, 1.15 to 1.34). In addition, repeated measures of BMI, WHR, and SMI had improved discriminatory power for predicting diabetes incidence compared to models considering only baseline measurements.
Results stratified by sex are shown in Fig. 1, Supplementary Tables 1 and 2. In men, for diabetes predictive power, the repeated measurements of SMI were highest Harrell’s C-index (0.662), followed by PBF and FMR (0.660), WC (0.656), and WHR (0.647). These did not differ from those of the baseline measures. In women, WHR had the highest C-index among repeated measures (0.691), followed by BMI (0.690) and WC (0.687). These were significantly improved compared to the predictive power of baseline measures, including FMR. In repeated measures, all indices with improved predictive power showed positive cNRI values, with BMI in the overall population and WHR in women showing the highest cNRI values (Supplementary Table 3). On the other hand, ABSI had the lowest C-index observed in both sexes regardless of repeated or baseline measurements. Additionally, in repeated measures, SMI in men contributed 7.1% to the Harrell’s C-index improvement, whereas in women WHR contributed 5.5%. However, these C-index was comparable to that of the remaining indicators except ABSI and BAI. In both sexes, all indices had independent effects on onset diabetes, even controlling for various diabetes-related features. The HR of WC in males and the HR of WHR in females showed the highest values among repeated measures (Supplementary Fig. 2). In women, when the predictive power of WC, BMI, WHR, and FMR was calculated by replacing measurements 8 years later, with values at 8 years (the mid-point of follow-up), the differences from the predictive power for time-varying measurements were no longer significant (Supplementary Fig. 3). Furthermore, the highest discriminatory power for predicting diabetes in premenopausal and postmenopausal women was observed with repeated measurements of WHR (0.726) and BMI (0.667) (Supplementary Table 4).
During the observation period, weight changes were more common in women than in men (36.2% and 26.7%). In a sensitivity analysis targeting individuals who experienced a weight change of ≥|10%| compared to baseline values, the highest C-index values were found in the repeated measurements of SMI in men and FMR in women, while ABSI, BAI, and BMI in men and ABSI in women did not differ from the basic model in model fit. Additionally, in women, the model applying WC, WHR, SMI, PBF, and FMR from repeated measurements improved predictive power compared to the baseline measurement model. But, in men, there was no improvement in the Harrell’s C-index (Supplementary Table 5).
Using 16 years of follow-up data, we found that anthropometric and body composition indicators were independently associated with incident diabetes and that their contribution to predicting incident diabetes varied by sex. In repeated measurements, the overall diabetes prediction power was greater than that of the model considering only baseline measurements, and the highest C-index was shown in WC in the entire population, SMI in men, and WHR in women. Notably, except for ABSI and BAI, the diabetes predictive power of the remaining indicators was comparable. Additionally, repeated measurements of WC, BMI, WHR, and FMR in women were found to contribute to improved discrimination compared to baseline measurements, but not in men. Interestingly, even in subjects who experienced weight change, repeated measurements of WC and WHR differed from baseline measurements for diabetes predictive power, primarily in women.
Consistent with prior studies [3,16], anthropometric and body composition parameters were found to have an independent effect on diabetes development. For precise predictions, repeated measurements were preferred to avoid bias due to measurement error, as a single measurement cannot encompass changes in measurements over time. However, our study showed comparability between baseline and repeated measures for diabetes prediction for some indices, which is supporting the conclusions of a recently performed meta-analysis [3]. This may have influenced the results due to the high reliability of intra-individual variability in anthropometric and body composition parameters. In this study, ICC of indices except ABSI also showed good reliability [15]. Studies using follow-up cohort data from middle-aged adults also showed flat patterns of anthropometric indicator trajectories [20,21]. Therefore, although a single measurement may be sufficient to estimate the general risk of incident diabetes, there may still be a risk of missing predictions in individuals with dynamic changes. Indeed, in sensitivity analyses, the discrimination power of repeated measures of WC, WHR, SMI, PBF, and FMR in predicting diabetes in women was increased compared to those of baseline measurements. On the other hand, the comparability of diabetes prediction power between baseline and repeated measurements was more pronounced in men.
Similar to our study, several studies have reported that anthropometric indicators predict diabetes more accurately in women than in men [22], although this is not universally agreed upon [3]. Research using the middle-aged population of the UK Biobank revealed more pronounced body size differences related to T2DM in women than in men. Peter et al. [23] suggested that due to gender differences in fat storage capacity, women need to gain more weight to develop susceptibility to insulin resistance and diabetes. While definitive conclusions have not been reached regarding gender differences in the impact of body size on diabetes risk, these differences are consistently attributed to variations in fat distribution between the sexes [3,22]. Indeed, it was noticeable that the predictive power of WHR varied depending on how many repeated measurements were considered in predicting diabetes in women (Supplementary Fig. 3). In subgroup analysis, repeated measurements of WHR, regardless of menopausal status, were found to be more predictive of diabetes than considering only baseline measurements. Furthermore, metabolic changes influenced by postmenopausal shifts in adipose tissue may also play a role in women [24].
Regarding repeated measures, a 9.3-year follow-up cohort study in Japan found that WC gain was linked to a higher risk of incident diabetes [25], while a 10-year follow-up study in China noted that changes in WHR were associated with the disease [26]. However, these studies only evaluated the impact of short-term body size changes on later diabetes development and did not consider changes at each time point. The longitudinal Atherosclerosis Risk in Communities (ARIC) study, which assessed diabetes risk using repeated measures of anthropometric indices, found that the predictive power of WC, BMI, WHR, and waist-to-height ratio was similar across both genders and races [16]. Our results support these findings but additionally show that repeated measurements of WC, BMI, WHR, and FMR in women offer better predictive power than baseline measurements alone.
Several meta-studies [3,27,28] have consistently found a positive linear association between BMI and diabetes. A recent systematic review study found a strong dose-response relationship, with the relative risk of diabetes increasing by 54% for each 1 SD increase in BMI [3], mirroring our results. Evidence from longitudinal studies suggests that BMI has the highest predictive power for developing diabetes among anthropometric indicators [16,29]. However, although BMI is commonly used to assess general obesity, it has been criticized for its inability to account for body composition, specifically body fat distribution, which may affect its reliability in health outcome assessments [30].
Accumulating research indicates a strong link between central obesity and glucose metabolism [31]. Although central obesity is often considered a more precise predictor of diabetes than general obesity, particularly in Asians, a meta-analysis revealed that the association between BMI and diabetes risk was comparable to that for WC across Asia, Europe, and the United States [31]. WHR is another metric used to assess central obesity and differentiate between upper and lower body fat distribution. Both WC and WHR, indicators of visceral and subcutaneous fat, have been consistently linked to diabetes in numerous studies. However, they do not seem to offer significant advantages over general obesity indices in predicting diabetes risk [29,31,32]. Interestingly, in repeated measures, combining BMI and WHR showed better improvement in predicting diabetes incidence (data not shown). Even in those with significant weight change, WC and WHR were associated with diabetes prediction, although only in women. This suggests the need to track abdominal obesity in clinical settings in addition to general obesity.
ABSI, introduced in 2012 as independent of BMI, exhibits a minimal correlation with it (r=–0.02). A systematic meta-study highlighted a 35% increase in diabetes risk per 1 SD increase in ABSI, although its limited variability has raised concerns about its clinical utility [33]. Consistent with other research [16,34], our study found that ABSI had the lowest discrimination power for diabetes in both sexes and its contribution to diabetes prediction significantly lagged behind that of WC, BMI, WHR, SMI, PBF, and FMR. Despite its positive correlation with WHR and WC—markers of central obesity—ABSI’s low reliability could explain its limited impact on diabetes prediction. A European prospective study suggested that combining ABSI with BMI improves predictions of mortality [33]. However, in our findings, while both ABSI and BMI independently affected diabetes development, incorporating ABSI into the model did not enhance the predictive accuracy for diabetes (Δ Harrell’s C-index 0.008), a result consistent across baseline measurements (data not shown).
With growing recognition of muscle’s role in chronic diseases, indices like those for sarcopenic obesity (SO) have emerged to assess its impact on conditions such as diabetes. Studies have identified SO as increasing diabetes risk by 1.38 times [35]. Although SMI, PBF, and FMR were closely related, they showed weaker correlations with indicators of central obesity. While adding WC to the SMI model slightly improved fit, the increase in predictive power for diabetes was minimal (Δ Harrell’s C-index 0.015, data not shown). Both skeletal muscle and fat mass are known to have a significant impact on glucose metabolism and insulin resistance [36,37], but we failed to distinguish between FMR and PBF. A retrospective study in Japan found glycemic control in T2DM patients correlated with skeletal muscle changes, but not with BMI changes [38]. Despite their clinical relevance, inconsistencies in measurement methods and definitions contribute to heterogeneity in research findings across studies [39,40]. Furthermore, it has been suggested that muscle quality may be more metabolically important than muscle quantity [41], which needs to be evaluated in the future.
There are several limitations as well as strengths of the study, so caution is needed when interpreting the results. Results were generated among middle-aged adults from two specific communities, limiting generalizability. There was no evaluation of fat distribution, such as visceral fat and subcutaneous fat, so detailed interpretation is limited. Potential bias from measurement errors may affect the results, so we also calculated the results using the bootstrap method. Nevertheless, because our results were obtained through a long-term longitudinal design of the study, they have strengths in explaining causal interpretations. We evaluated the reliability and discriminatory power of diabetes prediction based on repeated measurements to reflect the individual’s status at each time point. It can help to minimize bias from measurement error. Additionally, we compared models that used repeated measures to those using only baseline measurements to determine if they provided additional predictive information. Beyond commonly used adiposity-related anthropometric indicators, we also considered body composition indicators, enhancing the evidence in an area previously lacking robust data [3].
Taken together, anthropometric and body composition indices are associated with incident diabetes and have high internal reliability, but utilizing repeated measures to predict diabetes may be helpful in predicting hidden risk, especially in women. Additionally, in line with previous studies [16,42], general adiposity can be used to predict diabetes in community and clinical settings, but abdominal adiposity index may provide better information, so it is recommended to include assessment of abdominal obesity alongside general measures.
Supplementary materials related to this article can be found online at https://doi.org/10.4093/dmj.2024.0357.
SUPPLEMENTARY METHODS.
dmj-2024-0357-Supplementary-Methods.pdf
Supplementary Table 1.
Effect of anthropometric indices on the incident diabetes: men (n=2,877)
dmj-2024-0357-Supplementary-Table-1.pdf
Supplementary Table 2.
Effect of anthropometric indices on the incident diabetes: women (n=3,153)
dmj-2024-0357-Supplementary-Table-2.pdf
Supplementary Table 3.
Comparison of baseline and repeated measurements for anthropometric indices on the incident diabetes
dmj-2024-0357-Supplementary-Table-3.pdf
Supplementary Table 4.
Effect of anthropometric indices on the incident diabetes in women according to menopausal status
dmj-2024-0357-Supplementary-Table-4.pdf
Supplementary Table 5.
Effect of anthropometric indices on the incident diabetes among those who experienced a weight change of ≥|10%| from the baseline until the event was observed
dmj-2024-0357-Supplementary-Table-5.pdf
Supplementary Fig. 1.
Correlation coefficients for baseline (r) or time-varying measurements (rrm) between anthropometric and body composition indices. (A) Baseline measurements. (B) Repeated measurements. ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.
dmj-2024-0357-Supplementary-Fig-1.pdf
Supplementary Fig. 2.
Effect of repeated measurements of anthropometric and body composition indices on the incident diabetes by sex: (A) men and (B) women. Results were presented as hazard ratio (HR) and 95% confidence interval (CI) calculated through time-dependent Cox proportional hazard models. It was adjusted for age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline. ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.
dmj-2024-0357-Supplementary-Fig-2.pdf
Supplementary Fig. 3.
Comparison of the predictive power of diabetes risk for waist circumference (WC), body mass index (BMI), and waist-to-hip ratio (WHR) in female. The dots represent Harrell’s C-index, and width of the bar represents the 95% confidence interval. The predictive power of each anthropometric measure was obtained by adjusting for age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline. FMR, fat-to-muscle ratio; FU, follow-up. aIndividual measurements after 6 years of follow-up were replaced with measurements at 6 years of follow-up, bIndividual measurements after 8 years of follow-up were replaced with measurements at 8 years of follow-up, cStatistical difference (P<0.05) in Harrell C-index values compared to the statistical model for repeated measurements.
dmj-2024-0357-Supplementary-Fig-3.pdf

CONFLICTS OF INTEREST

No potential conflict of interest relevant to this article was reported.

AUTHOR CONTRIBUTIONS

Conception or design: H.A.L.

Acquisition, analysis, or interpretation of data: all authors.

Drafting the work or revising: H.A.L.

Final approval of the manuscript: all authors.

FUNDING

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2021R1A2C1003176). It had no role in the design, analysis or writing of this article.

Acknowledgements
This study was conducted with bioresources from National Biobank of Korea, the Korea Disease Control and Prevention Agency, Republic of Korea (KBN-2021-024).
Fig. 1.
Comparison of the predictive power of diabetes risk between repeated versus baseline measurements of anthropometric and body composition indices, stratified by sex: (A) men and (B) women. ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio. aThe statistical difference (P<0.05) between the Harrell’s C-index of the statistical model applying baseline and repeated measures.
dmj-2024-0357f1.jpg
dmj-2024-0357f2.jpg
Table 1.
Characteristics of study subject (n=6,030)
Characteristic Total Men (n=2,877, 47.7%) Women (n=3,153, 52.3%) P value
Age, yr 50.8±8.6 50.4±8.4 51.2±8.7 <0.001
Education level
 Did not graduate high school 3,045 (50.9) 1,086 (38.0) 1,959 (62.7) <0.001
 Graduated high school 2,029 (33.9) 1,091 (38.2) 938 (30.0)
 Some college or higher 912 (15.2) 683 (23.9) 229 (7.3)
Current smoking 1,466 (24.6) 1,363 (47.6) 103 (3.3) <0.001
Alcohol intake
 None 3,047 (52.0) 783 (28.1) 2,264 (73.7) <0.001
 Moderate 2,096 (35.7) 1,361 (48.8) 735 (23.9)
 Heavy 721 (12.3) 646 (23.2) 75 (2.4)
Physical activity
 None 2,004 (33.8) 922 (32.5) 1,082 (34.9) 0.020
 Moderate 792 (13.4) 361 (12.7) 431 (13.9)
 Heavy 3,138 (52.9) 1,554 (54.8) 1,584 (51.2)
Total energy, kcal/day 1,851.3 (1,547.7–2,210.3) 1,919.4 (1,633.5–2,278.6) 1,771.8 (1,471.1–2,139.2) <0.001
Dietary diversity score (range 0–12)
 ≤5 1,127 (19.2) 541 (19.4) 586 (19.1) 0.476
 6–8 2,963 (50.5) 1,427 (51.1) 1,536 (50.0)
 ≥9 1,774 (30.3) 823 (29.5) 951 (31.0)
C-reactive protein, mg/dL 0.1 (0.1–0.2) 0.1 (0.1–0.2) 0.1 (0.1–0.2) <0.001
Unweighted genetic risk score
 Missing 1,127 (18.7) 499 (17.3) 628 (19.9) 0.037
 Q1 (low) 1,389 (23.0) 671 (23.3) 718 (22.8)
 Q2 1,132 (18.8) 530 (18.4) 602 (19.1)
 Q3 1,525 (25.3) 739 (25.7) 786 (24.9)
 Q4 (high) 857 (14.2) 438 (15.2) 419 (13.3)
HbA1c, % 5.5±0.3 5.6±0.3 5.5±0.3 0.024
Glucose, mg/dL 83.0±9.0 85.1±9.6 81.2±7.9 <0.001
Anthropometric indices
 ABSI 0.1±0.0 0.1±0.0 0.1±0.0 <0.001
 BAI 28.4±4.1 25.5±2.6 31.1±3.4 <0.001
 WC, cm 81.6±8.6 83.1±7.5 80.2±9.4 <0.001
 BMI, kg/m2 24.4±3.0 24.2±2.9 24.7±3.2 <0.001
 WHR 0.9±0.1 0.9±0.1 0.9±0.1 <0.001
 Weight-adjusted SMI, % 69.2±6.8 74.2±4.6 64.6±5.0 <0.001
 PBF, % 26.7±7.1 21.5±4.9 31.4±5.2 <0.001
 FMR 0.4±0.1 0.3±0.1 0.5±0.1 <0.001

Values are presented as mean±standard deviation, number (%), or median (interquartile range).

HbA1c, glycosylated hemoglobin; ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.

Table 2.
Effect of baseline measurements of anthropometric indices on the incident diabetes (total n=6,030)
Anthropometric & body composition indices Prediction power
500 times Bootstrap
Effect estimate
Harrell’s C-index Lower 95% CI Upper 95% CI LR test Bootstrap-corrected C-index Lower 95% CI Upper 95% CI HR Lower 95% CI Upper 95% CI P value
Basic model 0.610 0.593 0.628 Ref 0.598 0.580 0.614 -
Basic model+ABSI 0.621 0.603 0.639 <0.001 0.609 0.592 0.627 1.222 1.137 1.314 <0.001
Basic model+BAI 0.634 0.616 0.652 <0.001 0.624 0.606 0.640 1.315 1.231 1.404 <0.001
Basic model+WC 0.661a,b 0.644b 0.679b <0.001 0.654a 0.635 0.671 1.571 1.468 1.681 <0.001
Basic model+BMI 0.657b 0.639b 0.675b <0.001 0.649 0.632 0.666 1.527 1.430 1.631 <0.001
Basic model+WHR 0.647 0.629 0.665 <0.001 0.638 0.618 0.655 1.476 1.376 1.583 <0.001
Basic model+weight-adjusted SMI 0.652b 0.634b 0.670b <0.001 0.643 0.625 0.661 0.667 0.621 0.716 <0.001
Basic model+PBF 0.653b 0.635b 0.671b <0.001 0.644 0.626 0.662 1.504 1.402 1.614 <0.001
Basic model+FMR 0.652b 0.634b 0.670b <0.001 0.644 0.626 0.661 1.468 1.374 1.567 <0.001

The basic model included sex, age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline. P for the LR test was presented as a Bonferroni-adjusted P value, taking into account the number of hypothesis tests.

CI, confidence interval; LR, likelihood ratio; HR, hazard ratio; ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.

a The highest Harrell’s C-index among the indices within the same column,

b The indices have a Bonferroni-adjusted P value greater than 0.05, indicating no significant difference between them within the same column.

Table 3.
Effect of repeated measurements of anthropometric indices on the incident diabetes (total n=6,030)
Anthropometric & body composition indices Prediction power
500 times Bootstrap
Effect estimate
Harrell’s C-index Lower 95% CI Upper 95% CI LR test Bootstrap-corrected C-index Lower 95% CI Upper 95% CI HR Lower 95% CI Upper 95% CI P value
Basic model 0.610 0.593 0.628 Ref 0.598 0.580 0.614 -
Basic model+ABSI 0.621 0.603 0.639 <0.001 0.609 0.591 0.626 1.238 1.147 1.336 <0.001
Basic model+BAI 0.633 0.615 0.651 <0.001 0.623 0.604 0.641 1.329 1.246 1.416 <0.001
Basic model+WC 0.668a,b 0.650b 0.686b <0.001 0.661a 0.643 0.679 1.630 1.526 1.740 <0.001
Basic model+BMI 0.665b 0.647b 0.683b <0.001 0.658 0.641 0.675 1.578 1.483 1.679 <0.001
Basic model+WHR 0.664b 0.646b 0.682b <0.001 0.655 0.637 0.673 1.442 1.377 1.511 <0.001
Basic model+weight-adjusted SMI 0.662b 0.644b 0.680b <0.001 0.654 0.637 0.672 0.636 0.594 0.680 <0.001
Basic model+PBF 0.661b 0.643b 0.679b <0.001 0.653 0.636 0.672 1.563 1.461 1.672 <0.001
Basic model+FMR 0.660b 0.642b 0.678b <0.001 0.653 0.636 0.671 1.507 1.418 1.601 <0.001

The basic model included sex, age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline. P for the LR test was presented as a Bonferroni-adjusted P value, taking into account the number of hypothesis tests.

CI, confidence interval; LR, likelihood ratio; HR, hazard ratio; ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.

a The highest Harrell’s C-index among the indices within the same column,

b The indices have a Bonferroni-adjusted P value greater than 0.05, indicating no significant difference between them within the same column.

  • 1. GBD 2021 Diabetes Collaborators. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet 2023;402:203-34.PubMedPMC
  • 2. NCD Risk Factor Collaboration (NCD-RisC). Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128·9 million children, adolescents, and adults. Lancet 2017;390:2627-42.PubMedPMC
  • 3. Jayedi A, Soltani S, Motlagh SZ, Emadi A, Shahinfar H, Moosavi H, et al. Anthropometric and adiposity indicators and risk of type 2 diabetes: systematic review and dose-response meta-analysis of cohort studies. BMJ 2022;376:e067516.ArticlePubMedPMC
  • 4. Yokokawa H, Fukuda H, Saita M, Goto K, Kaku T, Miyagami T, et al. An association between visceral or subcutaneous fat accumulation and diabetes mellitus among Japanese subjects. Diabetol Metab Syndr 2021;13:44.ArticlePubMedPMCPDF
  • 5. Park JH, Lee MY, Shin HK, Yoon KJ, Lee J, Park JH. Lower skeletal muscle mass is associated with diabetes and insulin resistance: a cross-sectional study. Diabetes Metab Res Rev 2023;39:e3681.PubMed
  • 6. Wang N, Sun Y, Zhang H, Chen C, Wang Y, Zhang J, et al. Total and regional fat-to-muscle mass ratio measured by bioelectrical impedance and risk of incident type 2 diabetes. J Cachexia Sarcopenia Muscle 2021;12:2154-62.ArticlePubMedPMCPDF
  • 7. Zhao Q, Yi X, Wang Z. Meta-analysis of the relationship between abdominal obesity and diabetic kidney disease in type 2 diabetic patients. Obes Facts 2021;14:338-45.ArticlePubMedPMCPDF
  • 8. Fu S, Zhang L, Xu J, Liu X, Zhu X. Association between abdominal obesity and diabetic retinopathy in patients with diabetes mellitus: a systematic review and meta-analysis. PLoS One 2023;18:e0279734.ArticlePubMedPMC
  • 9. Kim Y, Han BG; KoGES group. Cohort profile: the Korean Genome and Epidemiology Study (KoGES) consortium. Int J Epidemiol 2017;46:e20.ArticlePubMed
  • 10. Lee HA, Park H, Park B. Genetic predisposition, lifestyle inflammation score, food-based dietary inflammatory index, and the risk for incident diabetes: findings from the KoGES data. Nutr Metab Cardiovasc Dis 2024;34:642-50.ArticlePubMed
  • 11. Lee HA, Park H. Metabolically healthy obese individuals are still at high risk for diabetes: application of the marginal structural model. Diabetes Obes Metab 2024;26:431-40.ArticlePubMed
  • 12. Byrd DA, Judd SE, Flanders WD, Hartman TJ, Fedirko V, Bostick RM. Development and validation of novel dietary and lifestyle inflammation scores. J Nutr 2019;149:2206-18.ArticlePubMedPMCPDF
  • 13. Kim J, Kim M, Shin Y, Cho JH, Lee D, Kim Y. Association between dietary diversity score and metabolic syndrome in Korean adults: a community-based prospective cohort study. Nutrients 2022;14:5298.ArticlePubMedPMC
  • 14. Irimata K, Wakim P, Li X. Estimation of correlation coefficient in data with repeated measures. Available from: https://support.sas.com/resources/papers/proceedings18/2424-2018.pdf (cited 2024 Oct 18).
  • 15. Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med 2016;15:155-63.ArticlePubMedPMC
  • 16. Hardy DS, Stallings DT, Garvin JT, Xu H, Racette SB. Best anthropometric discriminators of incident type 2 diabetes among white and black adults: a longitudinal ARIC study. PLoS One 2017;12:e0168282.ArticlePubMedPMC
  • 17. Pencina MJ, D’Agostino RB Sr, Steyerberg EW. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med 2011;30:11-21.ArticlePubMed
  • 18. Therneau T, Atkinson E. The concordance statistic. Available from: https://cran.r-project.org/web/packages/survival/vignettes/concordance.pdf (cited 2024 Oct 18).
  • 19. Haines MS, Leong A, Porneala BC, Meigs JB, Miller KK. Association between muscle mass and diabetes prevalence independent of body fat distribution in adults under 50 years old. Nutr Diabetes 2022;12:29.ArticlePubMedPMCPDF
  • 20. Salmela J, Lallukka T, Mauramo E, Rahkonen O, Kanerva N. Body mass index trajectory-specific changes in economic circumstances: a person-oriented approach among midlife and ageing Finns. Int J Environ Res Public Health 2020;17:3668.ArticlePubMedPMC
  • 21. Jeon J, Jung KJ, Jee SH. Waist circumference trajectories and risk of type 2 diabetes mellitus in Korean population: the Korean genome and epidemiology study (KoGES). BMC Public Health 2019;19:741.ArticlePubMedPMCPDF
  • 22. Li T, Quan H, Zhang H, Lin L, Lin L, Ou Q, et al. Type 2 diabetes is more predictable in women than men by multiple anthropometric and biochemical measures. Sci Rep 2021;11:6062.ArticlePubMedPMCPDF
  • 23. Peters SA, Huxley RR, Woodward M. Sex differences in body anthropometry and composition in individuals with and without diabetes in the UK Biobank. BMJ Open 2016;6:e010007.ArticlePubMedPMC
  • 24. Abildgaard J, Ploug T, Al-Saoudi E, Wagner T, Thomsen C, Ewertsen C, et al. Changes in abdominal subcutaneous adipose tissue phenotype following menopause is associated with increased visceral fat mass. Sci Rep 2021;11:14750.ArticlePubMedPMCPDF
  • 25. Tatsumi Y, Watanabe M, Nakai M, Kokubo Y, Higashiyama A, Nishimura K, et al. Changes in waist circumference and the incidence of type 2 diabetes in community-dwelling men and women: the Suita Study. J Epidemiol 2015;25:489-95.ArticlePubMedPMC
  • 26. Ye M, Robson PJ, Eurich DT, Vena JE, Xu JY, Johnson JA. Anthropometric changes and risk of diabetes: are there sex differences? a longitudinal study of Alberta’s Tomorrow Project. BMJ Open 2019;9:e023829.ArticlePubMedPMC
  • 27. Hartemink N, Boshuizen HC, Nagelkerke NJ, Jacobs MA, van Houwelingen HC. Combining risk estimates from observational studies with different exposure cutpoints: a meta-analysis on body mass index and diabetes type 2. Am J Epidemiol 2006;163:1042-52.ArticlePubMed
  • 28. Kodama S, Horikawa C, Fujihara K, Heianza Y, Hirasawa R, Yachi Y, et al. Comparisons of the strength of associations with future type 2 diabetes risk among anthropometric obesity indicators, including waist-to-height ratio: a meta-analysis. Am J Epidemiol 2012;176:959-69.ArticlePubMed
  • 29. MacKay MF, Haffner SM, Wagenknecht LE, D’Agostino RB Jr, Hanley AJ. Prediction of type 2 diabetes using alternate anthropometric measures in a multi-ethnic cohort: the insulin resistance atherosclerosis study. Diabetes Care 2009;32:956-8.PubMedPMC
  • 30. Pischon T. BMI and mortality-time to revisit current recommendations for risk assessment. Am J Clin Nutr 2021;113:3-4.ArticlePubMedPDF
  • 31. Vazquez G, Duval S, Jacobs DR Jr, Silventoinen K. Comparison of body mass index, waist circumference, and waist/hip ratio in predicting incident diabetes: a meta-analysis. Epidemiol Rev 2007;29:115-28.ArticlePubMed
  • 32. Vera-Ponce VJ, Zuzunaga-Montoya FE, Loayza-Castro JA, Ramirez-Ortega AP, Torres-Malca JR, Garcia-Lara RA, et al. Diagnostic accuracy of anthropometric markers of obesity for prediabetes: a systematic review and meta-analysis. Int J Stat Med Res 2023;12:115-25.ArticlePDF
  • 33. Ji M, Zhang S, An R. Effectiveness of a body shape index (ABSI) in predicting chronic diseases and mortality: a systematic review and meta-analysis. Obes Rev 2018;19:737-59.ArticlePubMedPDF
  • 34. Fujita M, Sato Y, Nagashima K, Takahashi S, Hata A. Predictive power of a body shape index for development of diabetes, hypertension, and dyslipidemia in Japanese adults: a retrospective cohort study. PLoS One 2015;10:e0128972.ArticlePubMedPMC
  • 35. Khadra D, Itani L, Tannir H, Kreidieh D, El Masri D, El Ghoch M. Association between sarcopenic obesity and higher risk of type 2 diabetes in adults: a systematic review and meta-analysis. World J Diabetes 2019;10:311-23.ArticlePubMedPMC
  • 36. Deshmukh AS. Insulin-stimulated glucose uptake in healthy and insulin-resistant skeletal muscle. Horm Mol Biol Clin Investig 2016;26:13-24.ArticlePubMed
  • 37. Han TS, Al-Gindan YY, Govan L, Hankey CR, Lean ME. Associations of BMI, waist circumference, body fat, and skeletal muscle with type 2 diabetes in adults. Acta Diabetol 2019;56:947-54.ArticlePubMedPMCPDF
  • 38. Matsuura S, Nagata S, Shibazaki K, Uchida R, Imai Y, Shibata S, et al. Increased skeletal muscle mass index was involved in glycemic efficacy following diabetes treatment, and changes in fat mass index correlated with the changes in the lipid ratio in type 2 diabetes. J Diabetes Complications 2024;38:108717.ArticlePubMed
  • 39. Cruz-Jentoft AJ, Bahat G, Bauer J, Boirie Y, Bruyere O, Cederholm T, et al. Sarcopenia: revised European consensus on definition and diagnosis. Age Ageing 2019;48:16-31.ArticlePubMedPDF
  • 40. Chen LK, Woo J, Assantachai P, Auyeung TW, Chou MY, Iijima K, et al. Asian Working Group for Sarcopenia: 2019 consensus update on sarcopenia diagnosis and treatment. J Am Med Dir Assoc 2020;21:300-7.PubMed
  • 41. Volpato S, Bianchi L, Lauretani F, Lauretani F, Bandinelli S, Guralnik JM, et al. Role of muscle mass and muscle quality in the association between diabetes and gait speed. Diabetes Care 2012;35:1672-9.ArticlePubMedPMCPDF
  • 42. Christakoudi S, Tsilidis KK, Muller DC, Freisling H, Weiderpass E, Overvad K, et al. A body shape index (ABSI) achieves better mortality risk stratification than alternative indices of abdominal obesity: results from a large European cohort. Sci Rep 2020;10:14541.ArticlePubMedPMCPDF

Figure & Data

References

    Citations

    Citations to this article as recorded by  

      • PubReader PubReader
      • ePub LinkePub Link
      • Cite this Article
        Cite this Article
        export Copy Download
        Close
        Download Citation
        Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

        Format:
        • RIS — For EndNote, ProCite, RefWorks, and most other reference management software
        • BibTeX — For JabRef, BibDesk, and other BibTeX-specific software
        Include:
        • Citation for the content below
        Do Time-Dependent Repeated Measures of Anthropometric and Body Composition Indices Improve the Prediction of Incident Diabetes in the Cohort Study? Findings from a Community-Based Korean Genome and Epidemiology Study
        Close
      • XML DownloadXML Download
      Figure
      • 0
      • 1
      Do Time-Dependent Repeated Measures of Anthropometric and Body Composition Indices Improve the Prediction of Incident Diabetes in the Cohort Study? Findings from a Community-Based Korean Genome and Epidemiology Study
      Image Image
      Fig. 1. Comparison of the predictive power of diabetes risk between repeated versus baseline measurements of anthropometric and body composition indices, stratified by sex: (A) men and (B) women. ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio. aThe statistical difference (P<0.05) between the Harrell’s C-index of the statistical model applying baseline and repeated measures.
      Graphical abstract
      Do Time-Dependent Repeated Measures of Anthropometric and Body Composition Indices Improve the Prediction of Incident Diabetes in the Cohort Study? Findings from a Community-Based Korean Genome and Epidemiology Study
      Characteristic Total Men (n=2,877, 47.7%) Women (n=3,153, 52.3%) P value
      Age, yr 50.8±8.6 50.4±8.4 51.2±8.7 <0.001
      Education level
       Did not graduate high school 3,045 (50.9) 1,086 (38.0) 1,959 (62.7) <0.001
       Graduated high school 2,029 (33.9) 1,091 (38.2) 938 (30.0)
       Some college or higher 912 (15.2) 683 (23.9) 229 (7.3)
      Current smoking 1,466 (24.6) 1,363 (47.6) 103 (3.3) <0.001
      Alcohol intake
       None 3,047 (52.0) 783 (28.1) 2,264 (73.7) <0.001
       Moderate 2,096 (35.7) 1,361 (48.8) 735 (23.9)
       Heavy 721 (12.3) 646 (23.2) 75 (2.4)
      Physical activity
       None 2,004 (33.8) 922 (32.5) 1,082 (34.9) 0.020
       Moderate 792 (13.4) 361 (12.7) 431 (13.9)
       Heavy 3,138 (52.9) 1,554 (54.8) 1,584 (51.2)
      Total energy, kcal/day 1,851.3 (1,547.7–2,210.3) 1,919.4 (1,633.5–2,278.6) 1,771.8 (1,471.1–2,139.2) <0.001
      Dietary diversity score (range 0–12)
       ≤5 1,127 (19.2) 541 (19.4) 586 (19.1) 0.476
       6–8 2,963 (50.5) 1,427 (51.1) 1,536 (50.0)
       ≥9 1,774 (30.3) 823 (29.5) 951 (31.0)
      C-reactive protein, mg/dL 0.1 (0.1–0.2) 0.1 (0.1–0.2) 0.1 (0.1–0.2) <0.001
      Unweighted genetic risk score
       Missing 1,127 (18.7) 499 (17.3) 628 (19.9) 0.037
       Q1 (low) 1,389 (23.0) 671 (23.3) 718 (22.8)
       Q2 1,132 (18.8) 530 (18.4) 602 (19.1)
       Q3 1,525 (25.3) 739 (25.7) 786 (24.9)
       Q4 (high) 857 (14.2) 438 (15.2) 419 (13.3)
      HbA1c, % 5.5±0.3 5.6±0.3 5.5±0.3 0.024
      Glucose, mg/dL 83.0±9.0 85.1±9.6 81.2±7.9 <0.001
      Anthropometric indices
       ABSI 0.1±0.0 0.1±0.0 0.1±0.0 <0.001
       BAI 28.4±4.1 25.5±2.6 31.1±3.4 <0.001
       WC, cm 81.6±8.6 83.1±7.5 80.2±9.4 <0.001
       BMI, kg/m2 24.4±3.0 24.2±2.9 24.7±3.2 <0.001
       WHR 0.9±0.1 0.9±0.1 0.9±0.1 <0.001
       Weight-adjusted SMI, % 69.2±6.8 74.2±4.6 64.6±5.0 <0.001
       PBF, % 26.7±7.1 21.5±4.9 31.4±5.2 <0.001
       FMR 0.4±0.1 0.3±0.1 0.5±0.1 <0.001
      Anthropometric & body composition indices Prediction power
      500 times Bootstrap
      Effect estimate
      Harrell’s C-index Lower 95% CI Upper 95% CI LR test Bootstrap-corrected C-index Lower 95% CI Upper 95% CI HR Lower 95% CI Upper 95% CI P value
      Basic model 0.610 0.593 0.628 Ref 0.598 0.580 0.614 -
      Basic model+ABSI 0.621 0.603 0.639 <0.001 0.609 0.592 0.627 1.222 1.137 1.314 <0.001
      Basic model+BAI 0.634 0.616 0.652 <0.001 0.624 0.606 0.640 1.315 1.231 1.404 <0.001
      Basic model+WC 0.661a,b 0.644b 0.679b <0.001 0.654a 0.635 0.671 1.571 1.468 1.681 <0.001
      Basic model+BMI 0.657b 0.639b 0.675b <0.001 0.649 0.632 0.666 1.527 1.430 1.631 <0.001
      Basic model+WHR 0.647 0.629 0.665 <0.001 0.638 0.618 0.655 1.476 1.376 1.583 <0.001
      Basic model+weight-adjusted SMI 0.652b 0.634b 0.670b <0.001 0.643 0.625 0.661 0.667 0.621 0.716 <0.001
      Basic model+PBF 0.653b 0.635b 0.671b <0.001 0.644 0.626 0.662 1.504 1.402 1.614 <0.001
      Basic model+FMR 0.652b 0.634b 0.670b <0.001 0.644 0.626 0.661 1.468 1.374 1.567 <0.001
      Anthropometric & body composition indices Prediction power
      500 times Bootstrap
      Effect estimate
      Harrell’s C-index Lower 95% CI Upper 95% CI LR test Bootstrap-corrected C-index Lower 95% CI Upper 95% CI HR Lower 95% CI Upper 95% CI P value
      Basic model 0.610 0.593 0.628 Ref 0.598 0.580 0.614 -
      Basic model+ABSI 0.621 0.603 0.639 <0.001 0.609 0.591 0.626 1.238 1.147 1.336 <0.001
      Basic model+BAI 0.633 0.615 0.651 <0.001 0.623 0.604 0.641 1.329 1.246 1.416 <0.001
      Basic model+WC 0.668a,b 0.650b 0.686b <0.001 0.661a 0.643 0.679 1.630 1.526 1.740 <0.001
      Basic model+BMI 0.665b 0.647b 0.683b <0.001 0.658 0.641 0.675 1.578 1.483 1.679 <0.001
      Basic model+WHR 0.664b 0.646b 0.682b <0.001 0.655 0.637 0.673 1.442 1.377 1.511 <0.001
      Basic model+weight-adjusted SMI 0.662b 0.644b 0.680b <0.001 0.654 0.637 0.672 0.636 0.594 0.680 <0.001
      Basic model+PBF 0.661b 0.643b 0.679b <0.001 0.653 0.636 0.672 1.563 1.461 1.672 <0.001
      Basic model+FMR 0.660b 0.642b 0.678b <0.001 0.653 0.636 0.671 1.507 1.418 1.601 <0.001
      Table 1. Characteristics of study subject (n=6,030)

      Values are presented as mean±standard deviation, number (%), or median (interquartile range).

      HbA1c, glycosylated hemoglobin; ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.

      Table 2. Effect of baseline measurements of anthropometric indices on the incident diabetes (total n=6,030)

      The basic model included sex, age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline. P for the LR test was presented as a Bonferroni-adjusted P value, taking into account the number of hypothesis tests.

      CI, confidence interval; LR, likelihood ratio; HR, hazard ratio; ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.

      The highest Harrell’s C-index among the indices within the same column,

      The indices have a Bonferroni-adjusted P value greater than 0.05, indicating no significant difference between them within the same column.

      Table 3. Effect of repeated measurements of anthropometric indices on the incident diabetes (total n=6,030)

      The basic model included sex, age, education levels, physical activity, alcohol intake, current smoking, total energy intake, dietary diversity score, and log-transformed C-reactive protein levels, and quartiles of unweighted genetic risk score at baseline. P for the LR test was presented as a Bonferroni-adjusted P value, taking into account the number of hypothesis tests.

      CI, confidence interval; LR, likelihood ratio; HR, hazard ratio; ABSI, a body shape index; BAI, body adiposity index; WC, waist circumference; BMI, body mass index; WHR, waist-to-hip ratio; SMI, skeletal muscle index; PBF, percent body fat; FMR, fat-to-muscle ratio.

      The highest Harrell’s C-index among the indices within the same column,

      The indices have a Bonferroni-adjusted P value greater than 0.05, indicating no significant difference between them within the same column.

      Lee HA, Park H, Park B. Do Time-Dependent Repeated Measures of Anthropometric and Body Composition Indices Improve the Prediction of Incident Diabetes in the Cohort Study? Findings from a Community-Based Korean Genome and Epidemiology Study. Diabetes Metab J. 2024 Nov 15. doi: 10.4093/dmj.2024.0357. Epub ahead of print.
      Received: Jul 04, 2024; Accepted: Sep 07, 2024
      DOI: https://doi.org/10.4093/dmj.2024.0357.

      Diabetes Metab J : Diabetes & Metabolism Journal
      Close layer
      TOP