All subjects gave informed consent. The protocols and consent forms were approved by the University of California, San Francisco institutional review board and Clinical Research Center where the study was conducted.
Non-obese, non-diabetic, healthy subjects between the ages of 20 and 50 were recruited from the local population. A body mass index (BMI) cutoff of less than 27 was chosen due to the wide range of insulin sensitivity values with no correlation to BMI reported for this population
. The inclusion cutoff for Asian Americans was set lower at ≤ 25 because of the increased susceptibility for insulin resistance and type 2 diabetes at lower BMI values in this population
. Women were premenopausal. Individuals with diabetes, cardiovascular diseases, HIV and other active infections, thyroid disorders, epilepsy, cancer, hepatitis, cystic fibrosis, sickle cell disease, asthma or renal disease were excluded. Subjects taking glucocorticoids, adrenergic agonists, psychotropic drugs, diuretics, beta blockers, HMG CoA reductase inhibitors, or any other medications known to affect insulin sensitivity, carbohydrate metabolism, or lipid metabolism were excluded.
Exercise and general physical activity pattern were determined using the questionnaire developed by
. This questionnaire generates a physical-activity index score from 3 to 15 based on work, sport and leisure time energy expenditure, with each category scored from 1 to 5 (lowest to highest activity level). Subjects with scores greater than 10 (population mean approximately 8.3) were not be enrolled in the study.
Subjects underwent a dietary history at enrollment by the clinical research center nutritionist and were placed on a weight maintenance diet in order to avoid the confounding effect of weight loss on insulin sensitivity. Subjects could not be on nutritional supplements for at least three months prior to enrollment.
Oral glucose tolerance test
A fasting 75 g oral glucose tolerance test (OGTT) was performed. Subjects with impaired glucose tolerance or impaired fasting glucose were excluded.
Euglycemic hyperinsulinemic clamp
A euglycemic hyperinsulinemic clamp
 was performed at baseline and after 16 weeks treatment with chromium picolinate or placebo. A primed-continuous infusion of regular human insulin was administered at a rate of 40 mU/min/m2 body surface area for 120 minutes. This insulin infusion rate is sufficient to suppress hepatic glucose production in a normal non-obese non-diabetic population
[19, 20]. Bedside blood glucose levels were measured at 5 minute intervals and the glucose level was maintained at approximately basal level with a variable infusion of 20% glucose. Glucose disposal values (M/LBM/I) were calculated as mg glucose infused per min per kg lean body mass (LBM) during the steady state period between 90 and 120 minutes divided by steady state insulin (SSI) levels (in μU/ml x 100).
Blood and urine chemistry
Glucose was determined in whole blood by the glucose oxidase technique (Sigma). Insulin levels were measured by ELISA (Millipore, Billerica, MA).
The urine chromium levels were measured by atomic absorption spectrometry with graphite furnace atomization (AAGF) with Zeeman Background Correction and the serum chromium levels were measured by Inductively Coupled Plasma Mass Spectrometry (ICPMS) with Collision Cell Technology at Quest Diagnostics Nichols Institute (Chantilly VA). For AAGF, the sample was diluted with a “matrix modifier” that helped control the atomization of Chromium at a specific temperature. For ICPMS, the sample was diluted with a weak nitric acid solution. A linear calibration curve was obtained on blank samples and performed before and after the assays. Elevated values were repeated with a new sample set-up to check for contamination issues.
Measurements of body composition indices
Height was measured with a research center stadiometer. Body weight was recorded. Waist and hip circumferences were measured by a standardized protocol. Body composition was assessed by dual-energy X-ray absorptiometry (DXA).
Percutaneous muscle biopsies were obtained from the belly of the vastus lateralis at time of the euglycemic hyperinsulinemic clamp. Biopsies were obtained both prior to and after 120 minutes of insulin infusion on opposite legs. After local anesthesia, a 5 mm diameter Bergstrom needle was passed through a 7 mm skin incision and subcutaneous tissue, and then advanced approximately 2 cm beyond the muscle fascia. The biopsy (75-100 mg tissue) was obtained with applied suction. The incision was closed with steri-strips and firm pressure applied.
Muscle RNA preparation
RNA was isolated from frozen muscle tissue using the PureLink™ RNA mini kit with TRisol Reagent (Invitrogen, San Diego, CA). 50 mg of frozen muscle tissue was homogenized in 1 ml TRisol® Reagent using the Precellys 24™ Homogenizer (Omni International, Kennesaw, Georgia). Following the tissue homogenization, 0.2 ml chloroform was added to homogenate, shaken vigorously, incubated at room temperature for 2-3 minutes and microfuged at 12,000 g for 15 sec. 400 μl of the clear top layer was transferred to a microfuge tube, and equal parts of 70% ethanol were added to the tube. 700 μl was loaded onto the PureLink spin cartridge to purify the total RNA as directed in the PureLink RNA manual.
Human insulin signaling PCR array
The human insulin signaling pathway PCR array (PAHS-030A, SA Biosciences/Qiagen) was used to quantify gene expression in muscle biopsy preparations before and after chromium therapy in 8 subjects. This array profiles the expression of 84 genes coding for insulin receptor-associated proteins (including insulin and receptors, insulin-like growth factors and receptors, SH3/SH2 adapter protein); PI-3 kinase pathway proteins, MAPK pathway proteins; primary target proteins for insulin signaling; and target proteins for PPARγ. Each subject was measured in duplicates both before and after treatment. The PCR reaction was performed using manufacturer instructions. Plots of the two technical replicates against each other as well as hierarchical clustering confirmed that the reproducibility of the PCR array was very high. Five housekeeping genes were used: B2M (beta-2-microglobulin), HPRT1 (hypoxanthine phosphoribosyltransferase 1), RPL13A (ribosomal protein L13a), GAPDH (glyceraldehyde 3- phosphate dehydrogenase) and ACTB (beta-actin). Based on quality control analyses, we chose to normalize PCR cycle counts (Ct) of the target cDNAs such that all arrays have the same average Ct of HPRT1, RPL13A and ACTB. The normalized cycle counts (ΔCt) are averaged for each pair of duplicates (ΔCt
*) and gene expressions are calculated as 2-(ΔCt*).
Chromium or placebo treatment
After completion of baseline tests, subjects were randomized to chromium picolinate or placebo 500 μg twice daily for 16 weeks and the tests repeated. Both the investigators and the subjects were blinded. This daily dose of chromium picolinate was chosen because it was reported in one study that 1000 μg daily dose had greater efficacy than a 200 μg daily dose
. The chromium picolinate and placebo were supplied by Nutrition21 Inc. (Purchase, NY 10577). Adherence was assessed by pill count. We also measured fasting chromium levels in serum and spot urine at the end of the study.
Statistical analyses were performed using PASW 18.0 (SPSS Inc., Chicago, IL, USA). After data cleaning and checking distributions to assure that all scores met assumptions, key patient characteristics at baseline were compared across the chromium and control groups using chi-square and t-tests. Between-group differences from pre- to post-assessment on LBM M/I were examined with a repeated measures ANCOVA (RM-ANCOVA) analysis controlling for patient characteristics (age, gender, ethnicity, baseline BMI and triglycerides), followed by pair-wise post-hoc tests. Within the chromium treatment group only, the association between chromium absorption and change in insulin resistance was examined in a multiple regression controlling for the same background patient characteristics. Based on the wide range of urinary and serum chromium levels, and the assumption that high chromium levels reflect greater chromium absorption, patients within the chromium group were divided (median split) into a high and low chromium absorption group. A RM-ANCOVA analysis was repeated to test the group x time interaction on M LBM/ I between patients in the placebo group, low chromium absorbers, and high chromium absorbers. Logistic regression analyses were used to explore potential baseline patient measures as predictors of chromium absorption. Exploratory analyses to test whether pre-post changes in body weight or composition or lipids were associated with chromium levels were performed with ANCOVA.
The group sample size of subjects included in the final data analyses (14 active and 15 placebo) had adequate power (.80) to detect an effect of f=.30 or higher. This is equivalent to a medium/large to large effect on insulin sensitivity and had been previously reported with chromium picolinate therapy
Two analyses to test for differential gene expression in the insulin signaling pathway between before and after treatment were performed in “R”
A two-group unpaired Student’s t-test assuming equal group variances. This is also the test performed by the online ‘RT2 ProfilerT PCR Array Data Analysis’ tool provided by the manufacturer.
A two-group paired Student’s t-test assuming equal group variances (test not provided by manufacturer).