A case-control design was used in this investigation. Approval to recruit subjects to participate was granted by the Institutional Review Boards of the Jordan University of Science and Technology located in Irbid in the northern part of Jordan (IRB approval no., 249/2016). Signature on a consent form was a prerequisite for participation. Recruitment of study subjects took place between August 2016 and April 2017 at the clinics of King Abdullah University Hospital, a tertiary hospital affiliated with Jordan University of Science and Technology. This hospital is currently the largest hospital in the northern part of Jordan in terms of its capacity, number of staff members, and number of patients that visit the hospital on a daily basis. This hospital serves residents of the following governorates: Irbid, Ajloun, Jerash, and Mafraq, which together totaled one million inhabitants. The hospital has a bed capacity of 683 that can be increased to 800 beds in emergency cases.

Four hundred subjects with an age range of 35–70 years were invited to participate in this investigation. Two hundred subjects were previously diagnosed with T2DM by two endocrinologists according to the American Diabetes Association guidelines [3]. All the diabetes cases were recruited from the endocrinology clinics at King Abdullah University Hospital. The remaining 200 subjects did not have any symptoms of T2DM at the time of the investigation and were recruited from the same tertiary hospital though not necessarily from the endocrinology clinics. Absence of diabetes in the control arm of the study was assessed through the withdrawal of blood following an overnight fast and the measurement of fasting blood glucose (FBG) on two separate occasions. Pre-diabetes subjects with a repeated FBG of 100–125 mg/dL were excluded from participating in this study. Subjects with Cushing’s syndrome, polycystic ovarian syndrome, thyroid dysfunction, or hyperprolactinemia were also excluded from both arms of the study. Subjects who were taking BCAA supplements were also excluded from both arms of the study. Non-diabetes subjects were selected by matching body mass index (BMI) and age to diabetes patients. Diabetes patients were grouped into the following BMI categories: <25.0 kg/m², 25.0–29.9 kg/m², or ≥30.0 kg/m², and the following age categories: ≤40 years, 41–50 years, 51–60 years, and 61–70 years. Frequency matching was then performed during the recruitment of non-diabetes subjects.

The participants’ height (cm) and weight (kg) were recorded at the time of their visit to King Abdullah University Hospital. BMI was calculated based on the above measurements using the following equation: BMI=weight (kg)/height² (m²). The waist circumference (WC) was measured in centimeters (cm) using a tape measure positioned at the midpoint between the inferior costal margin and the superior border of the iliac crest with the measurement site being at the level of the umbilicus.

After a 12-hour fast, 10 mL of blood was withdrawn from the study subjects in a plain tube containing a clot activator (AFCO, Amman, Jordan). Serum was obtained from the samples by centrifuging at 3,000×g for 5 minutes. The serum samples were later used to measure glucose, total cholesterol, triglyceride, and BCAA levels.

Serum BCAA levels were measured using a commercially available kit purchased from Sigma Aldrich (St. Louis, MO, USA) with leucine as the standard. The assay involves a coupled enzymatic reaction to determine the levels of all three BCAAs, resulting in a colorimetric product. The absorbance was measured using the ELx800 Microplate Reader (BioTek Instruments, Winooski, VT, USA). A wavelength of 450 nm was used for the measurements. Additionally, the serum samples were submitted to the laboratories of King Abdullah University Hospital to measure glucose, total cholesterol, and triglyceride levels using an automated analyzer system (Roche Diagnostics, Mannheim, Germany).

Statistical analyses were performed using the IBM SPSS software ver. 22.0 (IBM Corp., Armonk, NY, USA). The Student t-test was used to test for significant differences in BCAA levels, age, sex (male or female), BMI, WC, and total cholesterol, triglyceride, or glucose levels between diabetes and non-diabetes subjects. Binomial logistic regression was used to test the association between BCAA levels and diabetes after including the following variables: age, sex, BMI, WC, and total cholesterol and triglyceride levels. A P-value of <0.05 was used as a cut-off for significance.

During the course of this study, 200 diabetes and 200 non-diabetes subjects were eligible. The percentage of females was higher than males among non-diabetes (58%) and diabetes (51.5%) subjects (Table 1); however, no significant differences existed in sex distribution between non-diabetes and diabetes subjects (P=0.218). Diabetes subjects were significantly older than non-diabetes (57.81±10.18 versus 52.24±10.28 years, P<0.0001) (Table 1). Furthermore, diabetes subjects had a higher BMI (30.82±5.35 versus 29.78±5.24 kg/m², P=0.067) and a significantly larger WC (109.59±12.66 versus 102.27±12.25 cm, P<0.0001) (Table 1).

The biochemical profile of the subjects showed that diabetes subjects had significantly higher levels of fasting serum BCAAs (469.00±117.22 versus 418.71±113.72 μmol/L, P<0.0001), glucose (167.67±82.59 versus 86.65±10.84 mg/dL, P<0.0001), cholesterol (208.93±69.78 versus 185.92±40.06 mg/dL, P<0.0001) and triglycerides (163.5±99.36 versus 140.82±105.24 mg/dL, P<0.0001) (Table 1).

Given that previous reports found that serum BCAA levels could be modified by BMI [13], sex [14], and WC [15], we wanted to test whether serum BCAA levels remained associated with diabetes in our study population following adjustment with the above variables. Therefore, we used binomial logistic regression and found that serum BCAA levels significantly increased diabetes risk (odds ratio [OR], 1.004; 95% confidence interval [CI], 1.001–1.006; P=0.003) (Table 2) even in the presence of the above variables in the model. Noteworthy, total serum cholesterol showed the strongest association with diabetes in our model (OR, 1.008; 95% CI, 1.003–1.013; P=0.002) (Table 2).