OBJECTIVE â To investigate the effects of metformin on glycemic control, ... casian, 10 Asian) were given metformin and placebo each for a 12-wk period in a ...
O R I G I N A L
A R T I C L E
N
IDDM subjects have a substantially increased risk of cardiovascular disease, which accounts for —5075% of all deaths in these patients (1). The excess risk of cardiovascular disease in diabetic subjects is unexplained by elevated levels of conventional risk factors (2). Several studies, both in diabetic and nondiabetic subjects, have demonstrated a relationship between hyperinsulinemia and certain risk factors for cardiovascular disease, such as elevated concentrations of serum triglyceride DINESH K. NAGI, MRCP JOHN S. YUDKIN, MD, FRCP (3,4) and PAI-1 (5) and low concentrations of HDL cholesterol. Asian-Indian OBJECTIVE — To investigate the effects of metformin on glycemic control, insulin resistance, and risk factors for cardiovascular disease in NIDDM subjects from two ethnic groups subjects in the U.K. exhibit hyperinsu(Caucasian and Asian) with different risks of cardiovascular disease. linemia and increased insulin resistance RESEARCH DESIGN AND METHODS— A total of 27 subjects with NIDDM (17 Cau- (6) that may be related to their excess casian, 10 Asian) were given metformin and placebo each for a 12-wk period in a randomized, risk of glucose intolerance (7) and perdouble-blind, placebo-controlled crossover study, and the dose was increased after 1 and 6 wk, up to a maximum of 850 mg three times a day. Insulin resistance, glycemic control, and haps also to excess risk of coronary heart cardiovascular risk factors were assessed before and after each treatment phase. The end of 12 disease (8,9). wk of metformin treatment was compared with the end of 12 wk of placebo treatment. Metformin, a biguanide antihyRESULTS — Metformin treatment was associated with significant improvement in FPG at 6 perglycemic agent, has been used for and 12 wk (mean difference at 12 wk, - 3 . 0 8 mM, 95% CI -4.12 to - 2 . 0 4 mM, 1 1 treatment of NIDDM subjects for >30 yr P < 0.0001) and MCR of glucose (median difference 0.40 ml • kg" • min" , interquartile range - 0 . 1 0 to 1.30 ml-kg" 1 -min" 1 , P = 0.036). 0-cell function calculated by HOMA in Europe and in Canada, but is curalso improved significantly (median difference 14%, interquartile range 7 to 23%, P < 0.001). rently unavailable for clinical use in the Total triglyceride (median difference —0.2 mM, interquartile range —0.6 to 0.1 mM, P = U.S. It is an effective drug in lowering 0.034), total cholesterol (mean difference -0.52 mM, 95% CI -0.83 to -0.22 mM, P = 0.002), and LDL cholesterol (mean difference -0.40 mM, 95% CI - 0 . 6 4 to -0.16 mM, P = both FPG and postprandial plasma glu0.002) fell significantly on metformin treatment, whereas no significant changes were ob- cose (10) and compares well in its effiserved in HDL cholesterol. PAI-1 activity fell significantly (mean difference —5.3 AU/ml, 95% cacy and safety with sulfonylureas CI —8.2 to —2.4 AU/ml, P = 0.001), but plasma fibrinogen concentrations and platelet (11,12), the only other oral agents curfunction, spontaneous or agonist induced, were unaffected. UAE was lower on metformin rently used for treatment of NIDDM subtreatment (median difference —2.4 |xg/min, interquartile range —4.4 to —0.2 |xg/min, P = 0.004), but metformin had no significant effect on BP. The effects of metformin on glycemic jects. The precise mechanism of the drug is not clear, but it is accepted that the control and cardiovascular risk factors were generally similar in the two ethnic groups. CONCLUSIONS — These findings indicate that metformin treatment improves glycemic con- antihyperglycemic effect of metformin is trol, and lowers insulin resistance and risk factors for cardiovascular disease, including PAI-1, and not mediated through increased insulin may therefore be useful in the long-term management of NIDDM subjects who have a highriskof concentrations (13). Metformin has been cardiovascular disease. shown to improve insulin sensitivity (14,15), although one study failed to FROM THK ACADEMIC UNIT OF DIABETES AND ENDOCRINOLOGY, WHITTINGTON HOSPITAL, LONDON, UNITED show any effect of metformin on insulinKINGDOM. stimulated glucose uptake (16). Various ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO DINESH K. NAGI, MRCP, NATIONAL INSTITUTES OF HF.AI.TH, N1DDK, 1 5 5 0 EAST INDIAN SCHOOL ROAD, PHOENIX, A Z , 8 5 0 1 4 . studies have evaluated the effects of this RECEIVED FOR PUBLICATION 19 MARCH 1 9 9 2 AND ACCEPTED IN REVISED FORM 16 JULY 1 9 9 2 . drug on several risk factors for cardioN I D D M , NON-INSUUN-DEPENDENT DIABETES MELUTUS; F P G , FASTING PLASMA GLUCOSE; C I , CONFIDENCE INTERVAL; vascular disease in normal and NIDDM MCR, METABOLIC: CLEARANCE RATE; H O M A , HOMEOSTASIS MODEL ASSESSMENT; LDL, LOW-DENSITY LIPOPROTEIN; HDL, subjects, but nearly all these studies were HIGH-DFNSITY UPOPROTEIN; P A I , PLASMINOGEN ACTIVATOR INHIBITOR; W H R , WAIST-TO-HIP RATIO; S T R , SUBSCAPULARTO-IRI(I:PS RATIO; BMI, BODY MASS INDEX; BP, BLOOD PRESSURE; DBP, DIASTOUC BLOOD PRESSURE; W H O , WORLD of limited duration, and the majority HEALTH ORGANIZATION; PRP, PLATELET RICH PLASMA; (3TBG, (3-THROMBOGLOBUUN; PF4, PLATELET FACTOR 4; CV, were not placebo-controlled, so the reCOEFFICIENT O F VARIATION; I R I , IMMUNOREACTIVE INSULIN; A U , ARBITRARY UNIT; V L D L , VERY-LOW-DENSITY LIPOPROTEIN; sults remain inconsistent (17). We have A N O V A , ANALYSIS O F VARIANCE; s B P , SYSTOLIC BLOOD PRESSURE; I D D M , INSULIN-DEPENDENT DIABETES MELLITUS; U A E RATE, URINARY AlBUMIN EXCRETION RATE; R , SPEARMANS RANK CORRELATION. performed a randomized, double-blind, placebo-controlled crossover study with
Effects of Metformin on Insulin Resistance, Risk Factors for Cardiovascular Disease, and Plasminogen Activator Inhibitor in NIPDM Subjects
A study of two ethnic groups
s
DIABETES
CARE,
VOLUME
16, NUMBER
4 , APRIL
1993
621
Metformin and cardiovascular risk factors
27 NIDDM subjects. The aims of this study were 1. To investigate the effect of metformin on glycemic control, insulin resistance, and hyperinsulinemia by using a new, highly specific assay for insulin, as conventional assays for insulin show substantial cross-reaction with proinsulin-like molecules, especially in Asian subjects (18,19), and to relate changes in glycemic control to those in insulin resistance. 2. To study the effects of metformin monotherapy on risk factors for cardiovascular disease and PA1-1 in NIDDM subjects and the relationship of changes in these risk factors to those in glycemic control and insulin resistance. 3. To determine whether the effects of metformin in two ethnic groups (Asian and Caucasian), with different degrees of insulin resistance and at different risks for cardiovascular disease, are different. RESEARCH DESIGN AND METHODS — All subjects had diabetes diagnosed by WHO criteria (20). The subjects represented a random sample of NIDDM subjects attending a diabetic clinic, and the study was approved by the ethical committee of Islington Health Authority. Subjects with clinical or electrocardiographic evidence of ischemic heart disease were excluded, as were subjects with a previous history of thromboembolic disease, clinical or biochemical evidence of renal or hepatic disease, untreated proliferative diabetic retinopathy, or albustix-positive proteinuria. Female patients of childbearing age and those unable to give fully informed consent also were excluded. After initial screening, oral hypoglycemic antidiabetic treatment was discontinued, and FPG concentrations were estimated after 2 wk. Subjects whose FPG was > 15 mM were excluded from entry into the study. Metformin was given for a total
622
period of 12 wk in a randomized, double-blind, placebo-controlled design, and the dose was increased stepwise from 850 mg once daily for 1 wk to 850 mg twice daily for 5 wk and to 850 mg three times daily for a further 6 wk. Baseline assessment took place on the day of inclusion in the study, and a similar assessment took place after 12 wk of therapy (phase 1). After a washout period of 2 wk, subjects were reassessed as at entry into the trial and crossed over to the alternate treatment (phase 2); they were reassessed finally at the end of phase 2. Subjects also were seen at 1, 2, and 6 wk into each treatment phase for a more limited assessment. Each subject was given a fresh pack of metformin or matched placebo (prepared and supplied by Lipha Pharmaceuticals, West Drayton, UK) on each visit, and the number of tablets returned were counted on each occasion to check for compliance. At each visit, patients were fully questioned about side effects related to metformin and symptoms related to hyperglycemia, and were asked to report any other symptoms. A total of 33 subjects (20 Caucasian, 13 Asian) were included in the study, and 27 subjects (17 Caucasian, 10 Asian) finished the trial successfully.
Clinical and biochemical investigations Each subject attended the clinical investigation ward of the diabetic unit at the Whittington Hospital between 0800 and 1000 after a 12-h fast. Weight and height were recorded with the patients wearing light clothing and without shoes; skin-fold measurements were performed with Holtain calipers (Holtain, Crosswell, UK), including three measurements each of subscapular and triceps skin fold. Waist and hip measurements were taken with a steel tape, the waist being measured at the point of maximum girth between the subcostal margin and the anterior superior iliac spine. Hip girth was recorded at the site of the greater trochanter. WHR, STR,
and BMI (weight/height in kg/m2) were calculated. After 30 min, resting BP was recorded twice in the right arm using a random zero sphygmomanometer (Hawksley, Lancing, UK), with dBP recorded at phase V Korotkoff sounds. The mean of the two BP readings was taken. Ankle and brachial sBPs were recorded with ultrasonic surface blood flow detectors (Vasculascope), and ankle-to-brachial sBP ratios were calculated. Each subject had an indwelling venous cannula placed in the antecubital fossa, and a venous blood sample was withdrawn for various measurements (plasma glucose, GHb, serum fructosamine, lipids, C-peptide, insulin, PAI-1, platelet studies in whole blood and PRP, (3TBG, PF4, and fibrinogen). Samples for C-peptide, insulin, PAI-1, (3TBG, and PF4 were cold-centrifuged (at 4°C) immediately and stored at —80°C. An insulin sensitivity test was performed by the modified Harano technique (21), during which each subject was given, for 150 min, 6 mg • kg" 1 • min" 1 of 20% glucose and 50 mU • kg" 1 • h" 1 of human soluble insulin (Actrapid, Novo Nordisk, Basingstoke, UK) in 100 ml of sodium chloride solution (154 mM), as separate infusions using Travenol Flowgard pumps (Baxter Health Care, Compton, Newbury, UK). Arterialized blood samples were withdrawn from the opposite arm every 5 min for the last 30 min (120-150) through an indwelling cannula. Steadystate plasma glucose was calculated from the mean of the seven samples taken from 120-150 min, and this reflected insulin resistance. MCR of glucose was determined by dividing the glucose infusion rate by steady-state plasma glucose. An estimate of changes in f$-cell function and insulin sensitivity on metformin treatment also was obtained by HOMA (22), which employs the fasting concentrations of glucose and insulin or glucose and C-peptide to derive estimates of these variables based on a mathematical model. Although it is recognized that absolute values of these estimates depend
DIABETES CARE, VOLUME 16,
NUMBER 4, APRIL
1993
Nagi and Yudkin
on the insulin assay used (23), the model is valid for comparing within-subject changes in the estimates between therapy periods. Moreover, the use of C-peptide concentrations with FPG concentrations in the model circumvents the problems of variability from using different insulin assays with different specificities. A value of 100% is taken as normal for (3-cell function and for insulin sensitivity. Plasma glucose. Plasma glucose concentrations were analyzed by glucose-oxidase method (Beckman Analyzer, Brea, CA). The intra-assay CV of the blood glucose assay was 2.2%. GHb. GHb was measured using electroendosmosis (Corning, Halstead, UK). The normal range was 6.5-8.5% with an intra-assay CV of 3% and interassay CV of 5%. IRI. IR1 was measured with a standard kit based on a polyclonal antibody assay (I 125 A19 human insulin RIA kit, Novo, Basingstoke, UK) (24). Because of our previous finding that this assay crossreacts with intact proinsulin and 32,33 split proinsulin (18), we refer to these results as IRI. The intra-assay CV was 5%, and the interassay CV was 9.6%. Insulin. Insulin concentrations also were measured with highly specific two-site monoclonal-antibody-based assays (25). The original method was modified to a microplate method that uses amplified, end-point, enzymoimmunometric assay and was validated against the original immunoradiometric assay. Insulin concentrations thus measured have been termed true insulin. The assay was sensitive to 2.0 pM with an intra-assay CV of < 10% and an interassay CV of < 15%. C-peptide. C-peptide was measured with a radioimmunoassay kit for human C-peptide (Antiserum K6, Novo) (26). The intra-assay CV was 4.7%, and the interassay CV was 8.8%. This assay has 75% cross-reactivity with human proinsulin, but because the usual concentrations of proinsulin in diabetic subjects are —4% of those of C-peptide, this cross-reaction is of little relevance.
DIABETES CARE, VOLUME
16,
NUMBER 4,
APRIL
PAI-1. PAI-1 was measured by a chromogenic assay (27) (Kabi Vitrum, Uxbridge, UK) using a microplate method. The intra-assay CV was 4.5%, and the interassay CV was 8.5%. Results are expressed as AU/ml (one unit of inhibitor being defined as the amount that inhibits one IU of tissue plasminogen activators). (JTBG and PF4. pTBG (Amersham, Aylesbury, UK) and PF4 (Abbott, Walsingham, UK) were determined with radioimmunoassay kits. To exclude the effect of in vitro platelet activation, data were analyzed only in subjects whose (3TBG-to-PF4 ratio was >3.0 (28). All samples from each visit were analyzed in the same batch. CVs within assay were 9.9% for (3TBG and 12.3% for PF4. Fibrinogen. Fibrinogen was analyzed by the Von Claus method (BCL, Lewes, UK). Plasma for these assays was stored at — 40°C, and samples were analyzed in the same batch. The intra-assay CV was 1.4%, and the interassay CV was 3%. Spontaneous platelet aggregation. Spontaneous platelet aggregation in whole blood was performed over a 60min period using an Ultra-Flo whole blood platelet-counting method (29). The number of platelets were counted at 0, 10, 20, 30 and 60 min. The platelet count from 10 to 60 min was expressed as percentage of the platelet count at 0 min. PRP platelet aggregation. PRP platelet aggregation studies were performed with four concentrations of A DP, as described previously (30), expressing sensitivity as EC50 (agonist concentration producing 50% of maximum aggregation). Total triglyceride. Total triglyceride was estimated by an enzymatic method with the Triglyceride PAP kit (Roche, Welwyn Garden City, UK). The inter-assay CV of this assay was 1.2%. Total cholesterol. Total cholesterol was measured by an enzymatic method (CHOD-PAP method) (Boehringer Mannheim, Lewis, UK). The inter-assay CV was 4%. HDL cholesterol was measured after precipitation of VLDL and LDL cholesterol by a magnesium chloride/ phosphotungstate reagent and with the
1993
same enzymatic method. The interassay CV was 5.4%. LDL cholesterol. LDL cholesterol was calculated using the Friedewald equation (31). Urine albumin. Urinary albumin was measured with a sensitive in-house ELISA technique. The UAE rate was calculated from a timed overnight urine collection. The interassay CV was