Design and study population

The data for this secondary cross-sectional study was obtained from the 2015 Colombian Health, Well-being and Ageing Survey (SABE 2015, from the Spanish: SAlud, Bienestar & Envejecimiento, 2015), a multicentre project conducted from 2014 to 2015 by the Pan-American Health Organisation and supported by the Epidemiological Office of the Ministry of Health and Social Protection of Colombia ( https://www.minsalud.gov.co/ ). A more detailed description of the survey methods has been provided elsewhere. The study included the Colombian population aged ≥60 years old, and the indicators were disaggregated by age range, sex, ethnicity, and socioeconomic level. The target population for SABE-Colombia included all adults aged 60 years old and above residing in households. Following conventional practice for population surveys, institutionalised persons (in prisons, jails, nursing homes, and long-term or dependent-care facilities) were excluded. Institutional review boards involved in developing the SABE-Colombia study (University of Caldas, ID protocol CBCS-021-14, and University of Valle, ID protocol 09-014 and O11-015) reviewed and approved the study protocol. Written informed consent was obtained from each individual before inclusion and completion of the examination. The relevant permissions and details are available from https://www.minsalud.gov.co/ . The study protocol of the secondary analysis was approved by the Human Subjects Committee at Pontificia Universidad Javeriana (ACTA ID 20/2017-2017/180, FM-CIE-0459-17) in accordance with the Declaration of Helsinki (World Medical Association) and Resolution 8430 of 1993, of the then Colombian Ministry of Health, on technical, scientific, and administrative standards for conducting research with humans.

This research involved a secondary analysis of the SABE observational study. The estimated sample size was 24,553 individuals, assuming an 80 % response from the target sample of 30,691 individuals. The original sample size (from 244 municipalities) was 23,694 older Colombians. In this sub-sample, two out of every five people were called to give a blood sample. A total of 3453 people were included in this analysis.

Anthropometric measurements

The research teams of the coordinating centres (Caldas and Valle universities, Colombia) trained the data collection staff to carry out both the face-to-face interviews and physical measurements. The anthropometry variables measured included height and body weight, which were measured using a portable stadiometer (SECA 213®, Hamburg, Germany) and an electronic scale (Kendall graduated platform scale), respectively. BMI was calculated as kg/m ² from the measured body weight (in kg) and height (in m). WC was measured using a soft measuring tape placed half-way between the lowest rib and top of the iliac crest, on bare skin or wearing light clothing. The WtHR was calculated as the ratio of WC (in cm) to height (in cm). The other anthropometric indexes (VAI, BRI, and C-Index) were calculated using the following formulas: VAI = males: [WC/39.68 + (1.88 x BMI)] x (TG/1.03) x (1.31/HDL); females: [WC/36.58+(1.89 x BMI)] x (TG/0.81) x (1.52/HDL), after Amato et al.; BRI = 364.2–365.5 [1 − π‐ ² WC ² (m) height− ² (m)] ^1/2 , after Thomas et al.; C-Index = 0.109 ^{− 1} WC (m) [weight (kg)/height (m)] ^{− 1/2} , after Valdez.

Variable definition

T2DM was defined as fasting plasma glucose ≥126 mg/dl (≥7.0 mmol/l). The blood samples were centrifuged for 10 min at 3000 rpm 30 min after sampling. All samples were delivered to a single central laboratory (Dinamica Laboratories, Bogotá, Colombia) for analysis within 24 h. Serum fasting glucose was analysed using enzymatic colorimetric methods (Olympus AU5200, Melville, NY, USA).

Co-variables

For lifestyle characteristics, personal habits regarding alcohol intake (participants were categorised as: teetotallers; those who drink less than one day per week; who drink two to six days a week; and who drink every day) and tobacco use (participants were categorised as: non-smokers, i.e., people who have never-smoked; those who currently smoke; and ex-smokers) were recorded. A “proxy physical activity” report was drawn up using the following questions: (i) “Have you regularly exercised, such as jogging or dancing, or performed rigorous physical activity at least three times a week for the past year?”; (ii) “Do you walk between 9 and 20 blocks (1.6 km) at least three times a week, without resting?”; (iii) “Do you walk 8 blocks (0.5 km) at least three times a week, without resting?”. Participants were considered physically active if they gave an affirmative response to two of the three questions. Socioeconomic status (SES) was determined on a scale of 1 to 6, based on housing stratum, with 1 representing the highest level of poverty and 6 the greatest wealth. This classification was developed by the National Government of Colombia and considers the physical characteristics of the dwellings, as well as their surroundings. Classification into one of the six strata was taken to approximate the hierarchical socioeconomic differences from poverty to wealth.

Statistical analysis

Baseline characteristics were presented as mean and standard deviation (SD) for continuous variables, and frequencies and percentages were calculated for categorical variables. The normality of the data was assessed using one-sample Kolmogorov-Smirnov tests. The differences between diabetic and non-diabetic participants were analysed using the unpaired Student’s T-test or Mann Whitney U-test for continuous variables, and the Chi-squared-test for categorical variables. The correlation between the indices and glucose was analysed using Pearson’s test. A receiver-operating characteristic (ROC) curve analysis was conducted to test the diagnostic performance of the six anthropometric indices (BMI, WC, WHtR, VAI, BRI, and C-Index) for predicting T2DM. The area under curve (AUC) calculation was based on the technique of DeLong et al. The optimal cut-off value for the anthropometric indices was based on the shortest distance and the highest Youden’s index value. Additionally, the effect size (ES) and odds ratio (OR) were calculated to find out the predictive magnitude. This magnitude was interpreted using trivial (<0.02), small (0.20 to <0.50), moderate (0.50 to <0.80) and large (≥0.80) values. The non-parametric approach of DeLong et al. was also used to compare the AUCs. All the analyses were executed using SPSS version 24.0 (SPSS, Chicago, IL, USA) and MedCalc Statistical Software version 18.2 (MedCalc Software bvba, Ostend, Belgium). All P values were two-sided and those less than 0.05 were considered to indicate statistical significance.