New compounds, the thiazolidinediones, have recently been developed as glucose-lowering agents. Early studies showed that the glucose-lowering effect of thiazolidinediones was evident in animal models of type 2 diabetes mellitus but not those of type 1 diabetes mellitus [5, 6], suggesting that some endogenous insulin secretion is needed for these agents to act. Troglitazone has been shown to decrease levels of not only plasma glucose and glycosylated hemoglobin [7-13] but also insulin and C-peptide. These observations, coupled with direct measures of whole-body insulin sensitivity in a small number of patients with type 2 diabetes mellitus [7], suggest that troglitazone exerts its major glucose-lowering effect by ameliorating insulin resistance. However, it is not clear whether troglitazone exerts its major insulin-sensitizing effect predominantly in the liver or in peripheral tissues. We studied this issue using detailed metabolic measurements in a large group of patients with type 2 diabetes mellitus.

Methods

Top Methods Results Discussion Author & Article Info References

This multicenter study was conducted at six sites: University of Chicago, Chicago, Illinois; University of Southern California, Los Angeles, California; University of Rochester, Rochester, New York; University of Pittsburgh, Pittsburgh, Pennsylvania; University of California, San Diego, San Diego, California; and Yale University, New Haven, Connecticut. Sample size was projected on the basis of study design, major end points, and standard power analysis. Each center enrolled patients while adhering to a common protocol with the same inclusion and exclusion criteria. At each center, patients gave written informed consent to participate in the study, which was approved by the respective university human investigation committees. All patients were studied in a 6-month, randomized, placebo-controlled, double-blind protocol. Patients were randomly assigned to treatment according to a blocked randomization code (block size, five) that was generated by a central computer. In each center, study personnel (executors of treatment assignment) and patients were blinded to the treatment code. Patients were consecutively assigned to treatments; equal numbers of troglitazone or matching placebo tablets were dispensed in a double-blind fashion.

Patients

Patients had to have type 2 diabetes mellitus according to the criteria of the National Diabetes Data Group [14], HbA_1c levels above the upper limit of normal, and fasting C-peptide levels of 0.49 nmol/L or greater. Therapy with oral antidiabetic medication was discontinued before randomization. Patients were excluded if they had clinically symptomatic heart disease, had had a vascular occlusive event in the previous 3 months, had had cancer in the past 5 years, had a serum creatinine level greater than 176.8 µmol/L, or had serum amino-transferase levels above the upper limit of normal.

Study Design

After medical screening, a 2-week wash-out period was allowed for discontinuation of therapy with oral antidiabetic medication in patients who were taking such medication. Metabolic studies were done before patients were randomly assigned to one of five treatment groups: 100, 200, 400, or 600 mg of troglitazone daily or placebo. At 6 months, follow-up metabolic studies were repeated 24 hours after patients received the last troglitazone or placebo tablet. At baseline and 6 months, patients were hospitalized and fasted overnight before a meal tolerance test (day 1) and a euglycemic-hyperinsulinemic clamp procedure (day 2) [15]. During the study, patients were prescribed a diet designed to maintain baseline body weight. Dietary assessment at the time of enrollment determined the patient's caloric needs [16]. The prescribed diet consisted of 50% carbohydrates, 34% fat (ratio of saturated fat to polyunsaturated fat, 1:4) and 16% protein. Patients were seen at monthly outpatient visits between the baseline and 6-month metabolic studies so that their clinical condition could be monitored.

Meal Tolerance Test

At approximately 7:00 a.m., patients were placed on bed rest and an intravenous catheter was inserted into an antecubital vein for blood sampling. A small volume of normal saline (0.9%) was infused to maintain patency. At approximately 8:00 a.m., patients ingested a liquid formula meal (Sustacal-HC [Mead Johnson & Co., Evansville, Indiana], which contained 33% of total daily caloric requirements); this was followed 4 hours later by an identical meal. Fasting blood samples were drawn, and additional samples were obtained every hour thereafter for 8 hours. Samples were processed immediately and stored at –80°C for measurement of serum levels of glucose, insulin, free fatty acids, and triglycerides and plasma levels of C-peptide. Fasting blood was also drawn for measurement of HbA_1c. After completing the test, patients received an evening meal according to their prescribed diet. They then fasted until the end of the euglycemic-hyperinsulinemic clamp procedure the following day. The intravenous line was left in situ for the clamp procedure.

Euglycemic-Hyperinsulinemic Clamp Procedure

At 6:00 a.m., a 4-hour primed (corrected for ambient fasting plasma glucose level), continuous (2 mg/m² body surface area per minute) infusion of [6,6- ²H]-glucose (di-deuterated glucose) isotope into the antecubital vein began. During the third hour of infusion, a retrograde cannula was inserted into a contralateral hand vein. The hand was warmed for sampling of arterialized venous blood. A small volume of normal saline (0.9%) was infused through the sampling cannula to maintain patency. Blood samples were drawn at 10-minute intervals during the final 40 minutes of the fourth hour for measurement of plasma glucose and insulin levels and glucose isotope enrichment. After 4 hours of isotope infusion, a two-step priming dose of insulin was administered (480 mU/m² per minute followed by 240 mU/m² per minute; each lasted 5 minutes); this was followed by a continuous infusion of insulin (120 mU/m² per minute) that lasted 300 minutes (total, 5 hours). The plasma glucose level was allowed to decrease to 5.5 mmol/L; exogenous glucose (dextrose, 20 g/100 mL of water enriched to approximately 2.5% with di-deuterated glucose) was then infused to maintain the plasma glucose level, measured every 5 minutes, at 5.5 mmol/L. The basal isotope infusion was stopped when the exogenous glucose infusion began.

Patients also received a continuous infusion of potassium (KCl and KPO₄), 0.105 mmol/L per minute, during the insulin infusion to maintain the serum potassium level between 3.5 and 4.5 mmol/L. During the final hour of the clamp procedure, blood samples were drawn every 10 minutes for measurement of plasma insulin levels and steady-state glucose isotope enrichment. For comparison with diabetic patients, eight persons without diabetes (mean age ±SD, 46 ± 6 years; mean fasting plasma glucose level, 5.3 ± 0.2 mmol/L; mean body mass index, 29 ± 3 kg/m²) were also studied on one occasion under basal and clamped conditions after an identical hyperinsulinemic clamp protocol.

Substrate and Hormone Measurements

Serum and plasma samples were shipped frozen to Corning Nichols Institute for chemical analysis and to Yale University for measurement of isotope enrichment. Serum total triglyceride levels (Boehringer Mannheim Diagnostics, Indianapolis, Indiana) and plasma free fatty acid levels (NEFA C-test, Wako Chemicals, Richmond, Virginia) were determined enzymatically; interassay coefficients of variation were 2% and 3.6%, and intraassay coefficients of variation were 1.6% and 1%, respectively. Insulin and C-peptide levels were measured by radioimmunoassay (Corning Nichols Institute); the interassay coefficients of variation were 12.3% and 12.0%, and the intraassay coefficients of variation were 7.4% and 6.5%, respectively. Levels of HbA_1c were measured by high-performance liquid chromatography using BioRad (Hercules, California) equipment (Corning Nichols Institute), with a normal reference range of 0.045 to 0.059. At each center, plasma glucose levels were measured at the bedside by using a Beckman glucose analyzer (Fullerton, California).

Glucose Isotope Data

Gas chromatography mass spectrometer analysis of enrichment of di-deuterated glucose in plasma and infusates was done at one center (Yale Stable Isotope Core Facility, New Haven, Connecticut) by using the penta-acetate derivative of glucose [17].

Calculations

Basal hepatic glucose production was calculated as follows:

Basal hepatic glucose production = (f/sa) x ([enrichment^inf/enrichment^plasma] –1), where f = basal [6,6- ²H] glucose infusion rate (mg/min), sa = body surface area (m²), enrichment^inf = [6,6- ²H] glucose infusate enrichment (%), and enrichment^plasma = steady-state basal plasma [6,6- ²H] glucose enrichment (%). The term enrichment refers to the fraction of isotope of glucose to naturally occurring (native) glucose, expressed as a percentage.

Clamped hepatic glucose production was calculated as follows:

Clamped hepatic glucose production = GIR x ([enrichment^inf/enrichment^plasma] –1), where GIR = mean glucose infusion rate (mg/m² per minute) during the clamp procedure (260 to 300 minutes), enrichment^inf = exogenous [6,6- ²H] glucose infusate enrichment (%), and enrichment^plasma = steady-state clamped plasma [6,6- ²H] glucose enrichment (%).

Clamped glucose disposal rate was calculated as follows:

Clamped glucose disposal rate = Clamped hepatic glucose production + GIR.

Statistical Analysis

Descriptive and inferential statistical analysis was done by using Systat, version 5.2.1 for Macintosh (SPS Inc., Chicago, Illinois). Comparisons over time were made by using repeated-measures analysis of variance. Post hoc comparisons done to localize effects were performed with simple analysis of variance and with the Dunnett and Tukey procedures for multiple comparisons. Forward stepwise regression analysis was used to determine the best models for predicting changes in fasting and postprandial plasma glucose levels. The independent variables considered in these models were fasting levels of insulin, C-peptide, free fatty acids, and triglycerides; basal glucose production; body mass index before treatment; and troglitazone dosage. All data in the text and tables are expressed as means ±SD. Mean differences with 95% CIs are also expressed in the text for key end points. For data from the meal tolerance test, the mean of hourly blood samples (from 1 to 8 hours inclusive) after the fasting blood sample was used as a measure of postprandial substrate and hormone levels.

Role of Funding Source

Funding authorities had no role in the analysis or interpretation of the data or in the subsequent decision to submit the report for publication.

Results

Top Methods Results Discussion Author & Article Info References

One hundred fourteen patients underwent baseline testing, and 93 completed the 6-month study. The 21 patients who did not complete the study were evenly distributed among the five treatment groups. Three patients had adverse events that led to early withdrawal from the study: One patient in the 200-mg/d group had a cerebrovascular event, 1 patient in the 200-mg/d group had appendicitis, and 1 patient in the 100-mg/d group had foot ulceration. Two patients were lost to follow-up during the study (1 patient each in the 100- and 600-mg/d groups), and another patient (in the 100-mg/d group) withdrew consent shortly after entering the study. Two patients (1 in the placebo group and 1 in the 100-mg/d group) did not comply with the protocol. For 2 additional patients (1 in the placebo group and 1 in the 100-mg/d group), data could not be interpreted because of errors made during the clamp procedure. Finally, 11 patients (6 in the placebo group, 3 in the 600-mg/d group, and 1 each in the 100- and 400-mg/d groups) withdrew from the study because of persisting hyperglycemic symptoms and documented worsening hyperglycemia.

Baseline morphometric and metabolic characteristics of the 93 patients who completed the study were similar among the six centers (Table 1) and among the five study groups (Table 2). None of the 93 patients had a serious adverse event or significant side effects of treatment, with the exception of one patient in the 400-mg/d group. This patient had a transiently elevated aminotransferase level at 6 months. After 6 months, significant weight loss occurred in the placebo group (loss of 3.3 ± 3.2 kg; P < 0.001); weight did not change from pretreatment levels in the troglitazone groups. Levels of HbA_1c increased during the study (F-ratio, 14.9; P < 0.001), but post hoc analysis showed that this change was significant only in the placebo group (baseline level, 0.08 ± 0.02; level at 6 months, 0.10 ± 0.03; P = 0.02) and that no change from baseline occurred in the troglitazone groups.

Meal Tolerance

During the baseline meal tolerance test, all groups experienced increases over fasting levels for the following variables: postprandial plasma glucose levels (Figure 1), triglyceride levels (fasting level, 2.35 ± 1.55 mmol/L; postprandial level, 3.44 ± 1.98 mmol/L), insulin levels (fasting level, 193 ± 114 pmol/L; postprandial level, 667 ± 473 pmol/L), and C-peptide levels (fasting level, 0.80 ± 0.37 nmol/L; postprandial level, 1.98 ± 0.92 nmol/L). In contrast, postprandial free fatty acid levels were lower than fasting levels (fasting level, 0.87 ± 0.41 mmol/L; postprandial level, 0.44 ± 0.59 mmol/L). After 6 months of treatment, fasting and postprandial plasma glucose levels slightly increased in the placebo group (Figure 1 and Figure 2, top). In contrast, all four troglitazone dosages decreased fasting (F-ratio, 5.7; P < 0.001) and postprandial (F-ratio, 3.2; P = 0.016) plasma glucose levels. In the 400- and 600-mg/d groups, glucose levels were decreased by approximately 20% under fasting conditions (400-mg/d group: mean change, –1.2 mmol/L [95% CI, –2.2 to –0.3 mmol/L]; 600-mg/d group: mean change, –2.5 mmol/L [CI, –3.9 to –0.9 mmol/L]) and under postprandial conditions (400-mg/d group: mean change, –2.8 mmol/L [CI, –4.5 to –1.1 mmol/L]; 600-mg/d group: mean change, –3.1 mmol/L [CI, –4.9 to –1.3 mmol/L]).

View larger version (23K): [in this window] [in a new window]   Figure 1. Plasma glucose profiles during meal tolerance test before treatment (top) and after treatment (bottom). To convert glucose values to SI units (mmol/L), multiply by 0.05551.

View larger version (16K): [in this window] [in a new window]   Figure 2. Mean changes in fasting (black ovals) and postprandial (white ovals) plasma levels of glucose (top), free fatty acid (top middle), triglyceride (bottom middle), and C-peptide (bottom). Changes are given as the level after treatment minus the level before treatment. Vertical bars represent 95% CIs. To convert glucose, triglyceride, and C-peptide values to SI units (mmol/L for glucose and triglyceride; nmol/L for C-peptide), multiply by 0.05551, 0.01129, and 0.331, respectively.

Free fatty acid levels were also decreased during fasting conditions (F-ratio, 2.6; P = 0.018) (Figure 2, top middle), but this finding was significant only in the 600-mg/d group (mean change, –0.44 mmol/L [CI, –0.71 to –0.17 mmol/L]). After 6 months, troglitazone decreased fasting (F-ratio, 6.3; P = 0.012) and postprandial (F-ratio, 7.7; P < 0.001) triglyceride levels independently of dosage (fasting conditions: mean change, –0.45 mmol/L [CI, –0.68 to –0.21 mmol/L]; postprandial conditions: mean change, –0.57 mmol/L [CI, –0.82 to –0.32 mmol/L]) (Figure 2, bottom middle). Plasma insulin levels were also decreased during fasting (F-ratio, 12.3; P < 0.001) and postprandial (F-ratio, 23.9; P < 0.001) conditions; post hoc analysis showed that changes were significant only in the 200-, 400-, and 600-mg/d groups (fasting conditions: mean change, –50 pmol/L [CI, –74 to –27 pmol/L]; postprandial conditions: mean change, –153 pmol/L [CI, –215 to –91 pmol/L]). Both fasting (F-ratio, 37.9; P < 0.001) and postprandial (F-ratio, 108.1; P < 0.001) C-peptide levels decreased after 6 months in all five groups (fasting conditions: mean change, –0.25 nmol/L [CI, –0.32 to –0.17 nmol/L]; postprandial conditions: mean change, –0.62 nmol/L [CI, –0.74 to –0.51 nmol/L]) (Figure 2, bottom).

Euglycemic-Hyperinsulinemic Clamp Procedure

Fasting (Table 2) and steady-state plasma glucose levels during the final stages of the baseline clamp procedure did not differ among the five groups. In addition, all groups had similar basal and clamped hepatic glucose production and clamped glucose disposal rates (Table 2). Baseline clamp data showed a strong correlation (r = 0.57; P < 0.001) between fasting plasma glucose levels and basal hepatic glucose production (Figure 3); this association had a better curvilinear (rather than linear) regression line fit. Although troglitazone decreased fasting plasma glucose after 6 months, steady-state plasma glucose levels were equal in all groups during the post-treatment clamp procedure. Basal hepatic glucose production increased after 6 months of treatment (F-ratio, 9.4; P = 0.003; mean change: placebo group, 15.4 ± 34.4 mg/m² per minute; 100-mg/d group, 8.2 ± 17.5 mg/m² per minute; 200-mg/d group, 10.3 ± 25.2 mg/m² per minute; 400-mg/d group, 7.7 ± 29.5 mg/m² per minute; 600-mg/d group, –3.3± 8.5 mg/m² per minute), but post hoc analysis showed that this finding was significant only in the placebo group (mean percentage change, 23% [CI, 2% to 44%]). Compared with changes in the placebo group, basal hepatic glucose production was relatively suppressed (P = 0.020) in the 600-mg/d group (difference, –18.7± 12.2 mg/m² per minute). Troglitazone did not modify insulin's suppressive effect on hepatic glucose production during the clamp procedure.

View larger version (21K): [in this window] [in a new window]   Figure 3. Plot of basal hepatic glucose production against fasting plasma glucose levels in 93 patients with type 2 diabetes mellitus (black squares) and 8 controls without diabetes (white squares). To convert glucose values to SI units (mmol/L), multiply by 0.05551.

In contrast to their effect on hepatic glucose production, all troglitazone dosages had a striking effect on insulin-mediated glucose disposal rates (Figure 4) (F-ratio, 4.3; P = 0.003). This effect was most noticeable in the 400- and 600-mg/d groups, in which glucose disposal rates were increased by approximately 45% above pretreatment levels (400-mg/d group: mean change, 112 mg/m² per minute [CI, 72 to 152 mg/m² per minute]; 600-mg/d group: mean change, 91 mg/m² per minute [CI, 67 to 115 mg/m² per minute]), even though steady-state plasma insulin levels at the baseline and 6-month clamp procedures were the same in all groups.

View larger version (16K): [in this window] [in a new window]   Figure 4. Mean changes in glucose disposal rate measured during the clamp procedure. Changes are given as the rate after treatment minus the rate before treatment. Steady-state glucose disposal rates increased in all troglitazone groups compared with placebo (P < 0.005). Vertical bars represent 95% CIs.

Predictors of Response to Treatment

Stepwise regression showed that troglitazone dosage, pretreatment fasting C-peptide level, and pretreatment body mass index were significant predictors of a favorable decline in plasma glucose over the course of treatment (as measured by percentage change during fasting and postprandial conditions). Troglitazone dosage was a strong predictor (fasting conditions, P < 0.001; postprandial conditions, P = 0.01); after we controlled for this effect, fasting C-peptide levels were the best predictor (fasting conditions, P = 0.016; postprandial conditions, P < 0.001). Troglitazone recipients in the lowest tertile of fasting C-peptide levels (<0.62 nmol/L) had reductions in fasting and postprandial glucose levels of 0.38 mmol/L and 0.94 mmol/L, respectively. Patients in the highest C-peptide tertile (>0.92 nmol/L) had mean reductions of 0.7 mmol/L and 2.9 mmol/L, respectively. Lower body mass index was also a significant predictor of a favorable glucose-lowering effect (fasting conditions, P = 0.05; postprandial conditions, P = 0.02).

Discussion

Top Methods Results Discussion Author & Article Info References

Our study shows the favorable metabolic effects of troglitazone monotherapy in patients with type 2 diabetes mellitus. Most important, troglitazone decreased both fasting and postprandial glucose levels by as much as approximately 20% at dosages of 400 to 600 mg/d. Data from euglycemic-hyperinsulinemic clamp procedures showed that troglitazone increased insulin-mediated glucose disposal in peripheral tissues but had a less obvious effect on the liver. The powerful effect of troglitazone on systemic glucose metabolism was accompanied by effects on lipid metabolism: a dose-independent decrease in triglyceride levels and a decrease in free fatty acid levels at the highest dosage.

Data from euglycemic-hyperinsulinemic clamp procedures show clearly that troglitazone monotherapy increases insulin-mediated rates of glucose disposal. Clamped glucose disposal rate is an experimental corrolate of the physiologic insulin-mediated clearance of glucose from the circulation that occurs after meals. We therefore observed that troglitazone's effect on insulin-mediated glucose disposal was accompanied by a troglitazone-mediated reduction in postprandial plasma glucose levels during the meal tolerance test. The manner in which troglitazone may facilitate such an effect on insulin-mediated glucose disposal is of interest. Troglitazone acts as a ligand for a nuclear receptor (the peroxisome proliferator activated receptor [18]), augmenting insulin action by enhancing postinsulin receptor signaling [19, 20]. It is argued that adipose tissue is the key insulin-sensitive tissue in which troglitazone action is most critical; Northern blot analysis of human tissues has shown that white adipose tissue has the greatest messenger RNA expression for the peroxisome proliferator activated receptor [18]. Yet, the two tissue sites that have the greatest direct impact on circulating glucose levels are skeletal muscle and the liver: Skeletal muscle is the major site for insulin-mediated clearance of glucose from the circulation, and the liver is the major site of endogenous glucose production. Adipose tissue, on the other hand, is thought to play a minor role in systemic glucose metabolism. We observed a striking increase in peripheral glucose disposal, undoubtedly caused by augmented insulin-mediated glucose uptake by skeletal muscle. Whether this is a direct effect of troglitazone on skeletal muscle or a secondary phenomenon is not clear. Research has shown that messenger RNA for the peroxisome proliferator activated receptor is found in skeletal muscle and liver, albeit in much smaller quantities than are found in adipose tissue [18]; thus, a direct effect of drug on skeletal muscle is possible. Evidence indicates that troglitazone acts directly on skeletal muscle in animals [19, 21]. However, the drug-mediated sensitization to insulin action in skeletal muscle may also be secondary to changes in lipid metabolism. Strong evidence suggests that muscle uses free fatty acids and glucose in a reciprocal fashion [22] and that elevations in circulating free fatty acid levels inhibit insulin-mediated glucose disposal in muscle [23, 24]. The reduction in circulating triglyceride and free fatty acid levels during troglitazone treatment may have mediated an opposite effect, increasing peripheral glucose disposal by skeletal muscle.

Of further interest is the effect of troglitazone on fasting plasma glucose levels and hepatic glucose production. Previous studies showed that fasting plasma glucose level is strongly correlated with basal hepatic glucose production [25-27]. Using an adjusted priming dose of tracer and a long period of tracer equilibration [28, 29], we accurately measured basal hepatic glucose production and noted a strong relation between fasting plasma glucose and hepatic glucose production. With this association in mind, we observed a troglitazone-mediated decrease in fasting plasma glucose levels that was most significant in the 600-mg/d group. This finding was accompanied by a relative suppression of basal hepatic glucose production at this same troglitazone dosage. At lower dosages (100 to 400 mg/d), troglitazone did not affect basal hepatic glucose production. Yet, it is also of note that troglitazone did decrease fasting blood glucose levels in these treatment groups. This finding may be explained by a drug-induced increase in glucose disposal in peripheral tissues, even under basal conditions.

The potential effect of troglitazone on the liver is supported by in vivo data in animals that showed a troglitazone-mediated reduction in hepatic glucose output [30, 31], inhibition of hepatic gluconeogenesis [32], and stimulation of hepatic glycogen synthase activity [33]. Consistent with these findings, in vitro data have shown that troglitazone augments glucose use in liver cell preparations [21]. On the other hand, the decrease in basal hepatic glucose production that we observed may be secondary to a troglitazone-mediated decrease in systemic free fatty acid levels. It has been reported that plasma free fatty acid levels are strongly correlated to hepatic glucose output [34], and acute systemic elevations of free fatty acids increase hepatic glucose production [35]. In a recent study of a small group of patients with type 2 diabetes mellitus who were receiving 400 mg of troglitazone per day for 6 to 12 weeks, Suter and colleagues [7] reported results that are similar to our findings. Under identical clamp conditions, they reported similar increments in peripheral glucose disposal rates with troglitazone. However, they noted that troglitazone reduced basal hepatic glucose production; we observed this in the 600-mg/d group but not in the 400-mg/d group. In support of the hypothesis that free fatty acids may be important in this setting, the patients in Suter and colleagues' study who were receiving 400 mg of troglitazone per day had a significant decrease in free fatty acid levels; we observed this in the 600-mg/d group but not in the 400-mg/d group.

All study groups (including the placebo group) had a similar suppression of hepatic glucose production during the clamp procedure; this finding suggests that troglitazone did not affect insulin's ability to suppress hepatic glucose production. However, we did use a supraphysiologic insulin stimulus that increased insulin's effect on peripheral glucose disposal; this magnified differences between treatment groups but also may have made it impossible to detect more subtle differences at the hepatic level. This issue has been addressed in rats under clamped conditions. In that study [31], a supraphysiologic dose of insulin similar to that used in our study could not discern an effect of troglitazone on insulin's suppressing action on hepatic glucose production. When a lower insulin dose was used, however, the authors reported that troglitazone did sensitize insulin's effect on hepatic glucose production [31].

The insulin-sensitizing action of troglitazone is also indicated by the overall decrease in circulating levels of insulin and C-peptide levels during both fasting and postprandial conditions. However, a similar decrease in C-peptide occurred in the placebo group. In this instance, two separate mechanisms may be at play. In the placebo group, in which fasting blood glucose and basal hepatic glucose production increased during the study, the decrease in fasting C-peptide levels could represent a worsening glucose toxic effect on ß-cell secretion [36]. At the other end of the spectrum (at the highest troglitazone dosage), the decrease in fasting C-peptide levels was accompanied by a decrease in fasting blood glucose levels and basal hepatic glucose production. This finding indicates that troglitazone had sensitized the liver to insulin or reduced some other factor that was driving hepatic glucose output without any effect on insulin signaling, resulting in a sparing of ß-cell secretory requirements.

Another possible explanation for the decrease in insulin and C-peptide levels may be the discontinuation of sulfonylurea therapy. Approximately two thirds of the patients entered the study after having recently discontinued sulfonylurea therapy, whereas one third were receiving diet therapy alone. However, we found no differences in metabolic characteristics before and after treatment between the two groups. It therefore seems unlikely that discontinuation of sulfonylurea therapy played a significant role in causing a decrease in insulin and C-peptide levels.

Previous studies have shown the effect of thiazolidinediones on lipid metabolism in insulin-resistant animals [5, 31, 37, 38] and humans [11, 12]. This effect occurs through a reduction in very-low-density lipoprotein cholesterol output from the liver and increased very-low-density lipoprotein cholesterol clearance in the periphery [38]. The same lipid-lowering effect was seen in the patients we studied, manifesting as a dosage-independent reduction in circulating levels of triglycerides and a decrease in free fatty acid levels that was most notable at the highest troglitazone dosage. Although we focused on the glucose-lowering effects of troglitazone, the lipid-lowering effects of this drug should not be understated. As discussed above, the troglitazone-mediated alterations in lipid metabolism may play an integral role in facilitating the changes seen in glucose metabolism and may offer an independent advantageous effect in the treatment of the dyslipidemia that occurs in patients with type 2 diabetes mellitus.

Earlier animal studies showed that metabolic responsiveness to thiazolidinediones is seen in models of type 2 as opposed to type 1 diabetes mellitus, indicating that endogenous insulin allows thiazolidinedione action. Our data support this finding; we identified factors that predicted a therapeutic glucose-lowering response as an aid for the appropriate use of troglitazone in a clinical setting. Baseline fasting C-peptide levels, an indication of ß-cell insulin secretory reserve (and an index of insulin resistance), was a positive indicator of a good glucose-lowering effect. Troglitazone is advocated for use in patients with established type 2 diabetes who receive more than 30 U of insulin per day. It could also be argued that troglitazone might have a favorable glucose-lowering response if used in combination with sulfonylurea agents when endogenous insulin levels are increased pharmacologically or if used as monotherapy when type 2 diabetes mellitus is at an early stage (during which significant insulin secretory reserve may still exist).

Although we noted that troglitazone monotherapy decreased plasma glucose levels (as measured during the meal tolerance test), we did not observe a significant decrease in HbA_1c levels, a more long-term measure of glycemia, from pretreatment levels. This finding can be explained by the fact that when troglitazone monotherapy began, most patients were undergoing a period of increasing hyperglycemia through "washout" from previous therapy with an antidiabetic agent. This manifested as an increase in HbA_1c levels from the time of recruitment to the time of randomization and a further increase in all groups that peaked after 3 months of treatment. Unlike the placebo group, in which HbA_1c did increase at 6 months from pretreatment levels, the troglitazone groups had HbA_1c levels that decreased to pretreatment levels; this finding suggests that troglitazone did indeed affect HbA_1c.

In summary, troglitazone monotherapy at a dosage of 400 to 600 mg/d decreased plasma glucose levels by as much as 20% in patients with type 2 diabetes mellitus. In a general sense, the magnitude of the glucose-lowering effect that we observed is similar to that reported for sulfonylurea agents and metformin. In a subset of patients with good insulin secretory reserve (as indicated by higher fasting C-peptide levels [>0.92 nmol/L]), the glucose-lowering effect was more pronounced and troglitazone monotherapy could therefore be advocated in such patients. Our data suggest that the glucose-lowering effect is primarily due to an increase in glucose clearance by peripheral tissues, with little effect on the liver except at the highest dosage (600 mg/d). Whether the metabolic changes we observed are a direct consequence of troglitazone stimulating glucose clearance by peripheral tissues (skeletal muscle in particular) or an indirect consequence of systemic alterations in lipid metabolism is not clear.

From Yale University, New Haven, Connecticut; University of Chicago, Chicago, Illinois; University of Southern California, Los Angeles, California; University of Rochester, Rochester, New York; University of Pittsburgh, Pittsburgh, Pennsylvania; University of California, San Diego, San Diego, California; and Parke-Davis, Ann Arbor, Michigan.

Dr. Buchanan: University of Southern California School of Medicine, Room 6602, LAC and USC Medical Center, 1200 North State Street, Los Angeles, CA 90033.

Dr. Burant: Department of Medicine, University of Chicago Medical Center, 5841 South Maryland Avenue, Chicago, IL 60637.

Dr. Gumbiner: National Institute for Fitness and Sports, 250 University Boulevard, Indianapolis, IN 46202.

Dr. Hsueh: Division of Diabetes, Hypertension and Nutrition, University of Southern California School of Medicine, Room 8250, General Hospital, 1200 North State Street, Los Angeles, CA 90033.

Dr. Kelley: Department of Endocrinology, University of Pittsburgh, E-1140 Biomedical Science Tower, Pittsburgh, PA 15261.

Dr. Nolan: Endocrine Department, St. James Hospital, Trinity College Dublin, Dublin 8, Ireland.

Dr. Olefsky: Endocrine/Metabolism Division, University of California, San Diego, Veterans Affairs Medical Center (111G), 3350 La Jolla Drive, San Diego, CA 92161-3267.

Dr. Polonsky: Section of Endocrinology, Department of Medicine, University of Chicago Medical Center, 5841 South Maryland Avenue, Room M267, Chicago, IL 60637.

Mr. Valiquett: Parke-Davis Pharmaceutical Research, 2800 Plymouth Road, Ann Arbor, MI 48105.