Dose-Response Effect of Pioglitazone on Insulin Sensitivity and Insulin Secretion in Type 2 Diabetes

A total of 58 patients with type 2 diabetes underwent 75-g OGTT before and after 26 weeks of pioglitazone therapy in a randomized, double-blind, placebo-control, multicenter clinical trial. To be eligible, patients were required to have HbA_1c ≥7.0%, fasting plasma glucose (FPG) ≥140 mg/dl, and fasting C-peptide >1 ng/ml. Patients who used insulin or who had unstable proliferative retinopathy, impaired liver function (aspartate aminotransferase or alanine aminotransferase >2.5 × upper limit of normal), impaired kidney function (serum creatinine >1.8 mg/dl), or anemia were excluded. Patients taking previous antidiabetic therapy (sulfonylureas or metformin) underwent a 6- to 8-week single-blind washout period before the baseline OGTT was performed. After the washout period, only patients with HbA_1c ≥7.0% were enrolled. Patients were randomized to one of five parallel treatment groups: pioglitazone 7.5, 15, 30, or 45 mg/day or placebo. During the double-blind period, patients were seen every 2 weeks for the first 6 weeks and every 4 weeks for the remaining 20 weeks. At 26 weeks, all subjects underwent repeat 75-g OGTT. To minimize the confounding effect of weight loss on metabolic changes, no specific dietary modifications were recommended during the study.

During the OGTT, plasma glucose (PG) concentration was measured at 0, 1, and 2 h by the hexokinase method (Hitachi 747-200 Analyzer; Roche, Indianapolis, IN). Plasma insulin concentration was measured at 0, 1, and 2 h during OGTT by an automated enzyme immunoassay (Tosoh AIA-1200; Tosoh Medicus, South San Francisco, CA). HbA_1c was measured by automated ion-exchange high-performance liquid chromatography (Bio-Rad Variant Analyzer; Bio-Rad Diagnostics, Hercules, CA). Fasting serum lipid levels (total cholesterol, HDL cholesterol, and triglycerides) were determined enzymatically (Hitachi 747 Analyzer; Roche, Indianapolis, IN). LDL cholesterol was calculated from the Friedewald equation.

The whole-body ISI was determined from the OGTT as follows (9):

where FPI=fasting plasma insulin (μU/ml), FPG=fasting plasma glucose (mg/dl), and P̅G̅ and P̅I̅ represent the mean plasma glucose and plasma insulin concentrations during OGTT. This index provides a composite measure of the combined effects of hyperinsulinemia plus hyperglycemia on muscle and hepatic glucose metabolism and is highly correlated with insulin sensitivity measured with the euglycemic insulin clamp technique (9). In the previous study (9), whole-body ISI was calculated from PG and insulin concentrations measured every 30 min during OGTT. In the present study, PG and insulin were measured every hour during the OGTT. Therefore, using the previously published data (9), we reanalyzed the relationship between insulin sensitivity obtained from euglycemic insulin clamp and ISI derived from PG and insulin concentrations at 0, 1, and 2 h during OGTT. A highly significant correlation between these parameters was observed in nondiabetic (r=0.74, P < 0.001) and type 2 diabetic (r=0.67, P < 0.001) individuals.

Hepatic insulin sensitivity can be estimated from the FPG and FPI as follows (9,10):

This equation is mathematically equivalent to the reduced formula of the homeostasis model assessment (HOMA) (10), where k=22.5 × 18, and the index of hepatic insulin sensitivity correlates closely with that measured directly with tritiated glucose (9).

Statistical calculations were performed using StatView for Windows software (version 5.0; SAS Institute, Cary, NC). Values before and after treatment within each group were analyzed using paired Student’s t test. Comparison between groups was performed using ANOVA with Bonferroni/Dunn post-hoc testing. Comparisons over time were made using repeated measures ANOVA. Pearson’s correlations between continuous variables were used as a measure of association. For the association between the five groups and changes in metabolic variables after treatment with pioglitazone or placebo, we also reevaluated the significance using Spearman’s correlation coefficient by rank. All data are presented as the mean± SEM. P < 0.05 was considered statistically significant.

Fifty-eight patients were randomized to receive placebo (n=11) or 7.5 (n=13), 15 (n=12), 30 (n=11), or 45 (n=11) mg/day of pioglitazone. At baseline, there were no significant differences in age, ethnicity, sex, BMI, HbA_1c, FPG, or insulin concentrations between the placebo and four pioglitazone-treated groups (Table 1). There tended to be more men than women in the pioglitazone-treated groups compared with the placebo group. Within the four pioglitazone-treated groups, the distribution of men and women was similar. It is known that men respond less well to thiazolidinediones than women because of their lower percentage of body fat (11). Plasma lipid levels were similar in the pioglitazone and placebo groups (Table 1).

After 26 weeks of treatment, there was a dose-dependent increase in body weight and BMI in the pioglitazone-treated groups (Table 1). The increments in body weight (4.5± 0.7 kg) and BMI (1.6± 0.3 kg/m ²) were significantly greater (P < 0.01) in the 45-mg/day pioglitazone group than in the placebo group. Plasma HbA_1c and FPG also demonstrated a dose-dependent decrease, with significant decrements in HbA_1c in the 15-, 30-, and 45-mg/day pioglitazone groups (P < 0.05–0.001 versus placebo) (Table 1). The increase in HbA_1c in the placebo group most likely reflects the lack of equilibration of HbA_1c during the 6- to 8-week washout period and, to a lesser extent, the progressive nature of type 2 diabetes. A highly significant inverse relationship was observed between change in body weight and change in HbA_1c (n=58, r=−0.527, P < 0.0001). Fasting plasma insulin concentration decreased slightly in all pioglitazone-treated groups, reaching significance in the 45-mg/day group (Table 1). In the 30- and 45-mg/day pioglitazone groups, plasma HDL cholesterol increased significantly from baseline, and plasma triglyceride concentration tended to decrease compared with placebo (Table 1). Plasma total and LDL cholesterol did not change significantly from baseline or from the placebo group.

During the baseline study, there were no statistically significant differences in FPG or FPI concentrations, the incremental area under the plasma glucose curve (ΔAUC glucose) and the incremental area under the plasma insulin curve (ΔAUC insulin) during the OGTT, the insulinogenic index (ΔAUC insulin/ΔAUC glucose) from 0 to 120 min, the hepatic insulin sensitivity index, or the whole-body ISI between the placebo and each pioglitazone group (Table 2). After 26 weeks of treatment with 30 and 45 mg/day of pioglitazone, the PG concentration during OGTT was significantly reduced from baseline (P < 0.01–0.05) and from placebo (P < 0.01–0.05) (Table 2). In both the 30- and 45-mg/day pioglitazone groups, the ΔAUC glucose was also significantly reduced compared with the pretreatment OGTT (Table 2). After 26 weeks, the incremental plasma insulin responses during the OGTT were similar in the placebo, 7.5-mg pioglitazone, and 15-mg/day pioglitazone groups. However, the ΔAUC insulin increased significantly in the 30-mg/day pioglitazone group (P=0.03 versus placebo) and the 45-mg/day pioglitazone group (P < 0.01 versus placebo and versus baseline). After 26 weeks of pioglitazone treatment, the insulinogenic index (ΔAUC insulin/ΔAUC glucose from 0 to 120 min) increased significantly in the 30-mg/day pioglitazone group (P < 0.05 versus baseline and P=0.07 versus placebo) and the 45-mg/day pioglitazone group (P ≤ 0.02 versus placebo and baseline) (Table 2). The hepatic ISI increased significantly versus placebo in the 15-, 30-, and 45-mg/day pioglitazone treatment groups (Table 2). The whole-body ISI increased in the 30- and 45-mg/day pioglitazone treatment groups compared with baseline and placebo (Table 2).

The relationships between pioglitazone dose and change in the selected variables during the 26-week pioglitazone treatment period are shown in Fig. 1. Using Pearson’s correlation coefficient, significant positive associations were observed between the pioglitazone dose and increases in body weight, plasma insulin response during OGTT, insulinogenic index, hepatic ISI, and whole-body ISI. Significant negative associations were observed between the pioglitazone dose and decrements in HbA_1c, FPG, and mean PG concentration during OGTT. Using Spearman’s correlation coefficient, we also examined the association by rank among the five groups (placebo and four pioglitazone groups) and the change in the same selected variables. Highly significant associations between the pioglitazone dose and change in each variable were also observed (Fig. 1; see P values within parentheses).

This dose-ranging study demonstrates that pioglitazone, given as monotherapy in doses of 30 and 45 mg/day, reduces FPG concentration and glucose excursion during OGTT. The decreases in fasting and post-OGTT PG were associated with decrements in HbA_1c of 0.8 and 1.6%, respectively, from baseline and 2.0 and 2.9%, respectively, from placebo. In the diabetic group treated with 15 mg/day of pioglitazone, there was a significant decrease in HbA_1c compared with placebo (1.3%, P < 0.05) but not compared with baseline (Table 1).

The ΔAUC glucose during OGTT decreased significantly from baseline in the 30- and 45-mg/day pioglitazone groups (Table 1). This beneficial effect of pioglitazone on postprandial hyperglycemia is quite distinct from sulfonylureas (12) and metformin (13), which exert their primary effect on the FPG concentration. Although the meglitinides reduce the postprandial glucose excursion, their effect on FPG is quite modest (14). The improvement in postprandial hyperglycemia after pioglitazone treatment could result from 1) increased insulin secretion, 2) enhanced tissue (peripheral and/or hepatic) sensitivity to insulin, or 3) an improvement in the combined effects of hyperglycemia plus hyperinsulinemia to promote glucose metabolism.

The effect of thiazolidinedione treatment on insulin secretion in patients with type 2 diabetes is controversial (1,11,15,16). Some studies have reported a decrease in plasma insulin response to a glucose challenge (1,11), whereas others have failed to observe any significant change (15,16). We found an increase in plasma insulin response during OGTT in diabetic patients treated with 30 and 45 mg/day of pioglitazone (Table 2). An increased or maintained plasma insulin response in the presence of a decreased PG concentration suggests an improvement in β-cell function. Consistent with this, pioglitazone (30- and 45-mg/day dose) increased the insulinogenic index (ΔAUC insulin/ΔAUC glucose) (Table 2). We believe that the variable effect of thiazolidinediones on the plasma insulin response to a glucose challenge is explained by two opposing effects: 1) a decrease in fasting and postprandial glucose concentrations, leading to reduced glucose toxicity (17); and 2) an increase in insulin sensitivity, leading to reduced insulin secretion (18).

Previous studies with thiazolidinediones in patients with type 2 diabetes concluded that their primary mechanism of action is related to improved insulin sensitivity (1,16,19,20). However, each of these insulin clamp studies used pharmacologic insulin infusion rates (80,120, or 300 mU · m ^–2 ·min ^–1), which produced supraphysiologic plasma insulin concentrations. In a recent study (8) using more physiologic plasma insulin concentrations, pioglitazone had only a very modest effect to enhance insulin sensitivity in patients with type 2 diabetes. In the present study, 26 weeks of pioglitazone treatment (30 and 45 mg/day) significantly increased the whole-body ISI compared with baseline and compared with placebo (Table 2). This improvement in whole-body ISI was observed with mean plasma insulin concentrations of 36–46 μU/ml (Table 2). It is important to note that there are significant differences between the OGTT and euglycemic insulin clamp, including the presence (OGTT) or absence (euglycemic insulin clamp) of hyperglycemia and the route of glucose administration (oral versus intravenous). Moreover, in contrast to the insulin clamp, the whole-body ISI during OGTT is greatly influenced by 1) glucose-mediated glucose uptake and the combined effects of hyperinsulinemia plus hyperglycemia to augment glucose disposal, 2) less effective suppression of hepatic glucose production (21), and 3) splanchnic uptake of 30–40% of the ingested glucose load (21). With regard to the latter, Kawamori et al. (22) have suggested that pioglitazone augments splanchnic glucose uptake after glucose ingestion. These considerations suggest that mechanisms, in addition to improved insulin sensitivity in muscle, contribute to the improvement in whole-body insulin sensitivity during OGTT after pioglitazone therapy.

From the dose-response standpoint, the insulin-sensitizing effect of pioglitazone is clearly evident at doses of 30 and 45 mg/day (Table 2). There was a tendency for whole-body insulin sensitivity to increase at 15 mg/day, but this did not reach statistical significance. Although whole-body insulin sensitivity and dose of pioglitazone were linearly related (Fig. 1), the threshold for a clinically significant insulin-sensitizing effect on glucose homeostasis is ∼15 mg/day of pioglitazone.

The product of basal hepatic glucose production (measured with tritiated glucose) and the FPI concentration provides a direct measure of hepatic insulin resistance under postabsorptive conditions, whereas the inverse provides a measure of hepatic insulin sensitivity (9). Because basal hepatic glucose production is closely correlated with FPG concentration (23), the inverse of the product of FPG and insulin concentrations provides an index of hepatic insulin sensitivity (9). The hepatic ISI increased in the 15-, 30-, and 45-mg/day pioglitazone groups (P < 0.05–0.01 versus placebo). Previous studies examining the effect of troglitazone on basal hepatic glucose production provided apparently contradictory results (1,16,20). In two studies, troglitazone reduced hepatic glucose production (1,20), whereas one study (16) found no effect of troglitazone on basal glucose output by the liver. However, these studies failed to relate the basal rate of hepatic glucose production to the fasting plasma insulin concentration. Because troglitazone treatment was associated with a decrease in FPI, a decreased or even unchanged rate of hepatic glucose production implies an improvement in hepatic insulin sensitivity.

In type 2 diabetic subjects, weight gain causes worsening of insulin resistance and deterioration of glycemic control. In the present study, pioglitazone treatment (15–45 mg/day) for 26 weeks was associated with mean weight gain of 2.0–4.5 kg, which was significantly and inversely correlated with decreases in HbA_1c (r=−0.527, P < 0.0001), FPG (r=−0.381, P=0.03), and mean PG concentration during OGTT (r=−0.462, P < 0.001). Improved glycemic control, despite weight gain, has been reported with other thiazolidinediones, including rosiglitazone and troglitazone (11,15). Although decreased glucosuria (associated with improved glycemic control) could contribute to the weight gain, this would not be expected to enhance insulin sensitivity. The seemingly paradoxical relationship between weight gain and improved glucose homeostasis/insulin sensitivity most likely is explained by the basic cellular mechanism of action of the thiazolidinediones, which exert their effects through the PPARγ (3,4). PPARγ receptors are found primarily in adipose tissue (3), and their activation causes preadipocytes to differentiate into mature small fat cells (3,4). PPARγ activation also induces key enzymes involved in lipogenesis in these newly formed adipocytes (3,4). Plasma free fatty acids provide the major source of lipid for triglyceride synthesis in these newly formed adipocytes (3). Although PPARγ receptors are present in both visceral and subcutaneous adipose tissue in humans, thiazolidinediones cause the differentiation of preadipocytes into mature adipocytes only in subcutaneous fat depots (24). This most likely explains the finding that thiazolidinedione-induced weight gain is associated with an increase in subcutaneous fat tissue and a decrease in visceral abdominal fat content (15,25). Because increased visceral fat is associated with insulin resistance (26), a reduction in visceral fat would be expected to lead to an enhancement in insulin sensitivity. Moreover, elevated plasma free fatty acid concentrations/oxidation are common in type 2 diabetic patients and cause insulin resistance in both liver and muscle (27,28). Because thiazolidinedione treatment consistently reduces plasma free fatty acid levels (11,20,29), this may provide another explanation for the improvement in insulin sensitivity despite weight gain.

Dose-response analyses demonstrated significant linear, positive associations between the pioglitazone dose and decrements in HbA_1c, FPG, and mean PG concentration during OGTT, as well as between pioglitazone dose and increases in body weight, ΔAUC insulin, insulinogenic index, and hepatic and whole-body ISI (Fig. 1). These significant associations were observed whether or not the placebo group was included in the analysis. Although we cannot determine the maximally effective dose of pioglitazone from the present study, it can be concluded that pioglitazone (at doses ranging from 15 to 45 mg/day) causes a dose-dependent decrease in fasting and postprandial PG concentrations through improvements in hepatic/whole-body insulin sensitivity and in β-cell function in type 2 diabetic patients. For most metabolic parameters, statistically significant changes from baseline and from placebo are evident with pioglitazone doses of 30 and 45 mg/day, with a tendency for the changes to become significant at 15 mg/day. Although some type 2 diabetic patients may respond to a pioglitazone dose of 15 mg/day when given as monotherapy, most will require ≥30 mg/day to see clinically significant effects on insulin sensitivity, enhanced β-cell function, and improved glucose homeostasis.

Support for the multicenter trial was provided by Takeda.

We thank Elva Chapa and Lorrie Albarado for expert secretarial assistance in preparing the manuscript.