Effects of Metformin and Rosiglitazone Monotherapy on Insulin-Mediated Hepatic Glucose Uptake and Their Relation to Visceral Fat in Type 2 Diabetes

We recruited 30 patients in whom type 2 diabetes had been diagnosed 1–3 years before the study (13) and who had never been treated with drugs. Patients were excluded if they had a fasting plasma glucose value of <6.1 mmol/l or >11.0 mmol/l after the run-in period. Patients with cardiovascular disease, blood pressure >160/100 mmHg, abnormal hepatic or renal function, proliferative retinopathy, anemia, or corticosteroid treatment were excluded. Two patients were taking lipid-lowering medications (statin). The study protocol was approved by the ethics committee of the Hospital District of Varsinais-Suomi. Written informed consent was obtained from all subjects.

The first part of the study consisted of a 4-week run-in period with written diet instructions. Then, patients were randomized to treatment with rosiglitazone (2 mg b.i.d. for 2 weeks, thereafter 4 mg b.i.d.), metformin (500 mg b.i.d. for 2 weeks, thereafter 1 g b.i.d.), or placebo for a 26-week double-blinded trial. Positron emission tomography (PET) studies, in combination with [¹⁸F]2-fluoro-2-deoxyglucose ([¹⁸F]FDG) and the insulin clamp technique (14), were performed before the treatment and at week 26 to assess insulin-mediated whole lean body mass glucose uptake (M), glucose uptake in liver (HGU) and abdominal adipose tissue (visceral fat glucose uptake [VFGU] and subcutaneous fat glucose uptake [SFGU]), and glucose influx rate-constants (K_i) in liver (hepatic K_i) and adipose tissue (visceral fat K_i and subcutaneous fat K_i). All PET studies were performed after an overnight fast.

An eight-ring ECAT 931/08-tomograph was used for image acquisition (Siemens/CTI, Knoxville, TN). A 5-min transmission scan was performed to correct subsequent emission scans for photon attenuation. Two catheters were inserted, one in an antecubital vein for infusion of glucose and insulin and for injection of [¹⁸F]FDG and one in the opposite radial artery for blood sampling. At 0 min, a primed-continuous infusion of insulin (40 mU · min⁻¹ · m⁻²) was started. The study for each subject consisted of a 150-min normoglycemic hyperinsulinemic period (14). At 90 min, [¹⁸F]FDG was injected, and consecutive 18-min dynamic scans (six × 180-s frame) of the abdominal area and of the liver were obtained at 20 min. Arterial plasma radioactivity was measured once during each time frame.

All data were corrected for dead time, decay, and measured photon attenuation and were reconstructed iteratively (15). Liver tissue time-activity curves were derived from large circular regions of interest (ROIs) placed on two to four planes in the right lobe of the liver. Adipose tissue ROIs were drawn on magnetic resonance imaging (MRI) images and copied into [¹⁸F]FDG images to cross-sectional slices from identical planes (12). Plasma and tissue time-activity curves were analyzed graphically to quantitate the fractional rate of tracer transport and phosphorylation (16), reflecting the glucose influx rate constant into the hepatic, visceral, and subcutaneous fat tissue components (K_i, ml · min⁻¹ · ml⁻¹). Graphical analysis has been previously evaluated for in vivo application to liver [¹⁸F]FDG-PET data (17,18). Rate-constant values were multiplied by steady-state plasma glucose concentrations achieved during the clamp to derive glucose uptake (μmol · min⁻¹ · ml⁻¹) (19,20). For adipose tissue, a lumped constant of 1.14 was used (12). Whole-body glucose uptake was calculated during the period of PET scanning and normalized for lean body mass (M). To account for interindividual differences in steady-state plasma insulin concentrations, all parameters of insulin-mediated glucose uptake were normalized to an insulin concentration of 600 pmol/l. Body fat content and lean mass were obtained with the impedance method. Posthepatic insulin clearance rate was computed during the steady state of the clamp (21).

The abdominal region was imaged with a 0.23 T Outlook GP (Marconi Medical Systems, Vantaa, Finland) magnetic resonance imager using a body coil. Transverse T1-weighted field echo images with a time repetition value of 170 ms and a time echo value of 4 ms were obtained with the same pixel size as PET images. The level of the mid-slice and the upper and lower border of the area imaged were determined as earlier described (12). Abdominal adipose tissue masses were measured at the level of intervertebral disc L2/L3. Fat volume was converted to weight using a tissue density of 0.9196 mg/ml. Biochemical analyses were performed as previously detailed (12).

Differences in paired data were evaluated using the Student’s paired t test for single repeated measurements. One-way ANOVA was used for unpaired group comparisons. To test for the effects of treatment, mixed multivariate models were set up in which treatment-induced changes in any given parameter were the dependent variable and treatment group and baseline values were independent variables. In these models, comparison between any two treatments was carried out by contrasts, and potential confounders (such as imbalances in sex group distribution) were controlled for by entering them as additional independent variables. Regression analyses were carried out according to standard techniques; to avoid the influence of the common denominator, values of glucose uptake and K_i were used without normalization to insulinemia. All data are presented as means ± SEM. Statistical significance was set at P ≤ 0.05.

At baseline, the three groups were well matched (Tables 1 and 2). At 26 weeks, BMI was significantly lower than at baseline in metformin-treated patients only. Fasting plasma glucose levels were significantly decreased after both rosiglitazone and metformin treatment in comparison with placebo. Metformin treatment was also associated with a significant drop in HbA_1c, whereas the corresponding change with rosiglitazone treatment fell just short of statistical significance (P = 0.12). Fasting plasma free fatty acid (FFA) levels were similar before and after either treatment. Fasting plasma insulin concentrations were lower after both rosiglitazone and metformin but not after placebo.

On the baseline clamp, there were no significant differences in steady-state plasma glucose, FFA, and insulin levels and in M and M normalized by the steady-state plasma insulin level of the clamp (M/I) across treatment groups (Table 3). Plasma FFA concentrations were significantly suppressed in all groups during the clamp. Percentage-wise, FFA suppression was higher after rosiglitazone than after placebo or metformin, whereas posthepatic insulin clearance was higher after metformin only.

Rosiglitazone and metformin both enhanced whole-body glucose uptake (both as M and M/I), whereas placebo had no significant effect on this variable. However, only the changes induced by rosiglitazone were significant when compared with placebo (Table 3). When controlling for the concurrent changes in BMI, the change in M averaged −1 ± 4 μmol · min⁻¹ · kg⁻¹ in metformin-treated patients (NS) and +9 ± 3 μmol · min⁻¹ · kg⁻¹ in rosiglitazone-treated patients (P = 0.01 vs. placebo).

By bioimpedance, changes in total fat mass averaged −2.9 ± 1.0 kg in metformin-treated patients (P < 0.001 vs. placebo) and −1.3 ± 1.0 kg in rosiglitazone-treated patients (NS). In the metformin group, placebo-subtracted MRI estimates of changes in subcutaneous and visceral depots were −0.55 ± 0.29 kg (P = 0.07 vs. placebo and P = 0.03 vs. rosiglitazone) and −0.11 ± 0.09 kg (NS), respectively. In the rosiglitazone group, the corresponding changes in fat depots were 0.11 ± 0.31 kg (NS) and −0.18 ± 0.09 (P = 0.06 vs. placebo) in the subcutaneous and visceral depots, respectively. After either metformin or rosiglitazone treatment, insulin-mediated VFGU was higher than with placebo; a smaller change in SFGU was observed, but it did not reach statistical significance, likely due to a sample size limitation (Fig. 1).

Both rosiglitazone and metformin treatments were associated with a significant increase in HGU (P = 0.018 and P = 0.001, respectively; Student paired t test for single groups); versus placebo, the change reached statistical significance in the metformin group when controlling for sex (placebo-subtracted values = +0.008 ± 0.004 μmol · min⁻¹ · kg⁻¹ · pmol/l⁻¹, P < 0.03, for metformin and +0.007 ± 0.004, P < 0.07, for rosiglitazone; sex-unadjusted levels of significance P = 0.06 and P = 0.2, respectively) (Fig. 1). In the whole dataset, sex-adjusted changes in HGU were negatively related to changes in HbA_1c; changes in HGU were positively associated with changes in VFGU and in percent FFA suppression (Fig. 2).

In the present study carried out in obese patients with mild type 2 diabetes, 26 weeks of metformin treatment (2 g/day) resulted in the expected lowering of fasting plasma glucose and HbA_1c levels; this was associated with a modest weight reduction that was largely accounted for by loss of subcutaneous fat. Rosiglitazone monotherapy (8 mg/day), on the other hand, resulted in a somewhat lesser reduction in fasting glycemia and HbA_1c as compared with metformin, with no significant change in body weight. The MRI estimates of regional fat depots were compatible with a reduction in subcutaneous fat in metformin-treated patients and in visceral fat (VF) in rosiglitazone-treated patients.

From the metabolic standpoint, metformin treatment was accompanied by a marginal increase in whole-body insulin sensitivity, most likely determined by weight and HbA_1c reduction. Rosiglitazone treatment led to a more pronounced improvement in whole-body insulin sensitivity, which included insulin-mediated glucose disposal and insulin suppressibility of lipolysis. This pattern of responses is typical of peroxisome proliferator-activated receptor-γ (PPAR-γ) agonists (10,11).

The novel finding is that both metformin and rosiglitazone monotherapy increased insulin-mediated HGU. The sex difference in the response to treatment may be related to the different body fat distribution, the ratio of visceral to subcutaneous fat mass being generally lower in women (in our series, 25 vs. 33%, P = 0.05). However, the small sample size and some imbalance in sex distribution among treatment groups prevents any firm conclusion on the mechanisms mediating the influence of sex on HGU response to treatment.

The clinical relevance of the finding of improved HGU stems from the observation that HGU, either insulin- or glucose-stimulated, is markedly impaired in patients with type 2 diabetes, and hepatic insulin resistance is likely to contribute to their hyperglycemia, both through inappropriately high glucose output and reduced glucose uptake (1,9). In the whole data set, the improvement in HGU was significantly related to the decrease in HbA_1c (Fig. 2), indicating that chronic glycemic control is a likely mechanism for the observed enhancement in HGU. Rosiglitazone and metformin were equally effective in increasing HGU, despite the fact that the former was more effective and more specific (i.e., independent of changes in body weight and glycemia) than the latter in enhancing insulin-mediated whole-body glucose disposal. This suggests that an overall improvement in insulin resistance may extend to HGU, but the contribution of changes in peripheral insulin sensitivity may have a small impact in comparison with the nonspecific effect of improved glycemic control.

Metformin acts on the liver to suppress gluconeogenesis—either by potentiating the effect of insulin or by reducing hepatic extraction of substrates—and glycogenolysis, and it decreases the activity of glucose-6-phosphatase (8,9). Additionally, metformin enhances insulin binding to hepatocytes of insulin-resistant mice, and it stimulates tyrosine kinase activity (8). Glucose transport and glycogen synthesis represent postreceptor targets of this drug (8). Recent data indicate that metformin modulated these enzymatic activities by upregulating AMP kinase, a key regulator of fuel sensing (22). Decreased hepatic gluconeogenesis and glycogenolysis and augmented glucose uptake have been attributed to thiazolidinediones (10,23). Thiazolidinediones have been suggested to affect the insulin signaling cascade and the activity of glycogen synthase. The evidence that PPAR-γ-responsive elements have been found in glucokinase-encoding genes is most intriguing (10). Thus, it is conceivable that the stimulatory effect on HGU seen in our patients taking either metformin or rosiglitazone represents direct, if different, actions of the two agents on liver metabolism.

The possibility that the observed effect of metformin or rosiglitazone on HGU were mediated by other metabolic changes is also plausible, as suggested by the data on adipose tissue glucose uptake. In our patients, treatment with either drug resulted in a significant enhancement of glucose uptake in visceral adipose tissue, whereas the corresponding changes in subcutaneous fat depots were smaller (and statistically not significant, likely due to the limited sample size). Increased glucose uptake in adipose tissue translates into increased reesterification of FFAs through augmented generation of glycerolphosphate from glycolysis. Increased FFA reesterification reduces net lipolysis and FFA delivery to the circulation; in turn, a reduced supply of FFA to the liver may be coupled with increased glucose uptake. Hepatic FFA oxidation promotes gluconeogenesis, thereby feeding indirect glycogen synthesis, and it might also directly regulate glycogen turnover, whereas circulating glucose provides the substrate for direct glycogen formation. Furthermore, a more pronounced effect of treatment on VF rather than subcutaneous fat is especially compatible with an improved HGU because of the preferential drainage of viscerally derived FFAs into the liver. This metabolic sequence may explain the direct association between changes in HGU and changes in VFGU on the one hand and suppression of lipolysis on the other (Fig. 2). The in vivo studies described herein lack the numerousness and measurement accuracy that would be required to provide conclusive evidence for the hypothetical sequence of events outlined above. The observed pattern of relationships is nevertheless strongly suggestive of a causal link between fat glucose metabolism, particularly in the visceral area, and hepatic glucose metabolism. In previous studies, increasing peripheral FFA levels with the use of intralipid infusion has led to contrasting findings on hepatic/splanchnic glucose uptake (24,25). The relationship between insulin resistance and VF mass is well established; removal of VF is followed by improved whole-body insulin sensitivity and decreased hepatic glucose production (26,27). Taken together with these experimental data, the current results may help explain why small weight changes, first involving VF mass, have unexpectedly profound metabolic consequences (26).

In conclusion, metformin or rosiglitazone monotherapy significantly increased HGU, thus providing evidence that improved HGU contributes to the therapeutic efficacy of these drugs in patients with type 2 diabetes. Direct drug effects together with improved glycemic control were likely determinants of the heightened HGU. Our data draw attention to the concept that VF insulin sensitivity may be a determinant of liver glucose metabolism in human type 2 diabetes.

The GSK corporation has provided funds to the University of Turku and Turku University Central Hospital in order to conduct PET studies on rosiglitazone to treat type 2 diabetes.