Patients enrolled were 18 years or older and were identified as having IGT, IFG, or metabolic syndrome.
IGT was diagnosed by American Diabetes Association guidelines (1
), which require the following 2 criteria: (a) a plasma glucose concentration 2 hours after consuming 75 g of glucose of at least 140 mg/dL, but below 200 mg/dL, and (b) a fasting plasma glucose (FPG) concentration less than 126 mg/dL.
IFG was diagnosed using the American Diabetes Association criteria of a FPG concentration of 100 mg/dL or greater, but less than 126 mg/dL (1
Metabolic syndrome was diagnosed using National Cholesterol Education Panel Adult Treatment Panel III criteria (12
), requiring the presence of 3 of the following 5 criteria: waist circumference greater than 102 cm in men or greater than 88 cm in women; triglyceride concentration 150 mg/dL or higher; high-density lipoprotein cholesterol concentration less than 40 mg/dL in men or less than 50 mg/dL in women; blood pressure greater than 130/greater than 85 mm Hg; and FPG concentration 100 mg/dL or higher.
Patients were excluded if they were diabetic (FPG concentration greater than 126 mg/dL; 2-hour plasma glucose concentration greater than 200 mg/dL). Other exclusion criteria included self-reported hospitalization for treatment of cardiovascular disease 6 months before enrollment, serum creatinine concentration greater than 2.0 mg/dL at baseline, self-reported pancreatitis, recent or significant abdominal surgery, pregnancy and/or intention to become pregnant during the study, polycystic ovarian syndrome or irregular menses, or use of chromium supplements less than 1 month before screening.
Patients were excluded if they were taking drugs thought to affect glucose metabolism and/or endothelial function (glucocorticoids, antineoplastic agents, psychoactive agents, and bronchodilators). Patients taking antihypertensive drugs and lipid-lowering agents were allowed to participate provided that dosages were stable for 3 months before enrollment.
Ethical and Safety Considerations
The study protocol and consent form were approved by the Griffin Hospital (Derby, Connecticut) Institutional Review Board and the Yale University (New Haven, Connecticut) Human Investigation Committee and the study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained, and all participants received a total of $475 (divided among assessment visits) as compensation for their participation. For safety monitoring, unblinded subject treatment assignment was maintained by a data and safety monitoring board.
Study Design and Interventions
This study was a randomized, double-blind, placebo controlled, modified cross-over clinical trial to investigate the effects of daily chromium supplementation for 6 months at 2 dosage levels (500 mcg and 1000 mcg of chromium picolinate daily) on serum measures of insulin sensitivity and glucose tolerance in adults with IGT, IFG, and metabolic syndrome.
The study used a modified cross-over (Latin square) design encompassing both paired (cross-over) and unpaired comparisons with statistical methods and sample size tailored to serve both purposes (see Statistical Analysis).
The study was designed and powered to compare the 6-month effects of 500 mcg to 1000 mcg of chromium on insulin resistance (homeostasis model assessment of insulin resistance [HOMA-IR]), which provided adequate power (see Statistical Analysis) to detect a change in 2-hour plasma glucose and endothelial function. Effects of chromium at each dosage were compared with placebo as a paired (cross-over) comparison after 6 months of use. Because the time required for chromium to wash out fully is unknown, a posttreatment phase of 6 months was incorporated into the design following 12 months of intervention and placebo ().
Study Design and Flow Diagram.
Patients meeting eligibility criteria were randomly assigned to 500 mcg daily or 1000 mcg daily of chromium picolinate and then further randomized to chromium/placebo or placebo/chromium sequences. After completing the initial 6-month period, participants immediately crossed over to the alternate assignment (). All investigators and participants were blinded to treatment assignment.
The 2 dosages of chromium picolinate (500 mcg or 1000 mcg daily) and placebo capsules came in the form of capsules similar in shape, size, and appearance, donated by Nutrition 21, Inc (Purchase, New York). Placebo capsules were indistinguishable from those containing chromium. Supplying pharmacy personnel encoded the treatment supplements and matching placebos.
The primary outcome measures were serum insulin, HOMA-IR, 2-hour plasma glucose, fasting plasma glucose, and 2-hour insulin during oral glucose tolerance testing assessed after 6 months of chromium use. Insulin resistance was calculated with the HOMA-IR by using the following equation: HOMA-IR = fasting plasma insulin (µIU/mL) × FPG (mmol/L) × 22.5 (13
Secondary outcome measures included weight, waist circumference, body mass index (BMI), blood pressure, endothelial function as assessed by flow-mediated dilatation (FMD) of the brachial artery, blood glycohemoglobin, total serum cholesterol, serum high-density lipoprotein cholesterol, serum low-density lipoprotein cholesterol, serum triglycerides, and urinary microalbumin-to-creatinine ratio.
Participants were weighed after removing their heavy outer garments and shoes, standing on a calibrated scale, standing in the center of the platform with weight distributed evenly to both feet. Waist circumference was measured at the level of the umbilicus in a horizontal position. BMI was calculated at each visit. Blood pressure was determined using the Datascope Accutorr Plus automatic digital blood pressure device with the participant supine.
Our endothelial function assessment methods have been previously reported (14
). In brief, endothelial function as FMD was measured noninvasively in the right brachial artery with high-frequency ultrasonography (Sonos 4500; Phillips Medical Systems, Andover, Massachusetts) in accordance with published guidelines (15
). Measures of vessel diameter and flow velocity were obtained by a single dedicated vascular clinical research specialist blinded to subject treatment status. To account for potential variability in stimulus strength, FMD was divided by flow at 15 seconds after cuff deflation to create a stimulus-adjusted response measure. Endothelial function was assessed in the fasting state, with a nitroglycerin control, and at 90 minutes post-OGTT to assess the change in FMD between fasting and fed states. (A glucose load has been associated with transient endothelial dysfunction in healthy, prediabetic, and diabetic populations [16–19]; nutritional interventions [14,20,21] have been shown to mitigate acute [immediate] endothelial dysfunction.)
Laboratory measures including FPG, 2-hour plasma glucose, glycohemoglobin, lipid panel, and urinary markers were collected and analyzed at Griffin Hospital (Derby, Connecticut) using standard procedures at each visit. Insulin concentrations were measured at the Yale Center for Clinical Investigation Core Laboratory (New Haven, Connecticut).
A sample size of 60 participants, allowing for 20% attrition and nonadherence, was predicted to provide 90% power (8
) to detect a minimal difference of 9.5% in HOMA sensitivity between the 500-mcg and 1000-mcg arms (α = 0.05). A standard deviation of 10.1 was used on the basis of previous literature (22
). This sample size also provided greater than 80% power to detect a change of 15% in 2-hour plasma glucose concentration and 0.5% in FMD between the 500-mcg and 1000-mcg groups at 6 months.
Analysis was by intention-to-treat; 6-month postintervention analysis was conducted on those having completed 6-month assessments (primary end point) in the cross-over design. Missing individual data were addressed by the last observation carried forward method.
Statistical analysis was conducted using SAS software (Version 9.1, SAS Institute, Cary, North Carolina). Independent samples t tests, chi-square tests, and Mann-Whitney tests were used to compare participants at baseline. Descriptive and exploratory analyses of all measured outcomes were performed before embarking on modeling or hypothesis testing procedures. Normally distributed data were analyzed using repeated measures analysis of variance to assess differences in intraindividual responses across treatments for all outcome measures, while non-normally distributed data were analyzed using the Mann-Whitney test. Assessment of treatment effects between dosage groups, where baseline between-group differences were observed to be either statistically significant (P<.05) or approaching trend-level significance (P<.10), was conducted using analysis of covariance, with baseline values entered into the model to control for those differences. The combined effect of independent variables (age, sex, BMI at baseline, metabolic syndrome, and treatment sequence) and treatment assignment on all outcome measures was assessed with multivariable models.
To assess the effect of different treatment assignments on outcome measures, we computed 95% confidence intervals (CI) for mean changes from baseline following each treatment assignment. Comparisons across treatment assignments were made using the 95% CI around the mean change of outcome measures from their baseline values. When the 95% CI of mean change of one treatment assignment was not included within the 95% CI of mean change of another treatment assignment (ie, nonoverlap), we considered the 2 treatment assignments significantly different at P<.05. Mean change from baseline values of outcome measures after each treatment assignment was considered significant when the 95% CI around the mean change did not include zero.