Patients with type 2 diabetes mellitus exhibit an increased propensity to develop extensive arteriosclerosis with its sequelae, unstable angina pectoris and acute myocardial infarction (1- 2). Over the last years, experimental data have illuminated the role of inflammation in atherogenesis, while clinical studies have shown that this concept of inflammation in arteriosclerosis applies directly to human patients (3). As such, increased serum levels of inflammatory biomarkers of arteriosclerosis, like C-reactive protein, cytokines, like tumor necrosis factor-alpha or interleukin-6, as well as novel markers like monocyte chemoattractant protein (MCP)-1, soluble CD40 ligand (sCD40L), and matrix metalloproteinases (MMP) have been shown to predict cardiovascular risk and seem to reflect the overall burden of vascular disease in patients. Interestingly, some of these markers are elevated in patients with type 2 diabetes and insulin resistance, indicating a pivotal role of inflammation in this metabolic disorder (4- 6). Recent data suggest that the release of inflammatory mediators like tumor necrosis factor-alpha and interleukin-6 from the visceral adipose tissue as well as an activation of vascular cells itself contribute to the inflammatory state in these patients with metabolic syndrome. Therefore, enhanced serum levels of sCD40L may reflect endothelial and platelet activation in diabetic subjects, while increased MMP-9 levels suggest the presence of unstable plaques with activated monocytes/macrophages (7- 11). Moreover, elevated soluble adhesion molecules like soluble intracellular adhesion molecule (sICAM) and soluble vascular cell adhesion molecule (sVCAM) are markers of endothelial dysfunction in these patients. Given the increased risk of diabetic patients for macrovascular events, therapeutic strategies that limit inflammation in the vessel wall and reduce serum levels of inflammatory surrogate parameters have been considered a promising tool to influence vascular disease in this high-risk population (12- 13).

Recent experimental and clinical data suggest that a novel group of antidiabetic agents, thiazolidinediones (TZDs), like pioglitazone and rosiglitazone, may exhibit anti-inflammatory properties in the vessel wall (14- 15). These agents act via the nuclear transcription factor peroxisome proliferator-activated receptor gamma (PPARγ) and, in addition to their metabolic action, have been shown to regulate the expression of various target genes in vascular cells in vitro and in vivo, subsequently limiting inflammatory cell activation and lesion formation during atherogenesis. Furthermore, clinical data suggest that TZDs reduce inflammatory biomarkers of arteriosclerosis, like C-reactive protein or sCD40L, in treated patients, thus potentially modulating their cardiovascular risk (16- 19). Still, most of these studies were placebo-controlled and, as such, did not allow the dissection of metabolic from nonmetabolic TZDs’ effects, because TZD treatment, compared to placebo, significantly improved glucose metabolism in all of these studies. Therefore, we performed a six-month prospective, randomized, controlled trial to compare the effect of pioglitazone and sulfonylurea treatment on inflammatory biomarkers of arteriosclerosis, attempting to achieve comparable improvement in blood glucose control in both treatment groups.

The prospective randomized monocentric study was approved by the responsible ethics committee and was performed in accordance with the Declaration of Helsinki and the guidelines for good clinical practice. All patient examinations were performed at the Clinical Department of the Institute for Clinical Research and Development (IKFE), Mainz, Germany. After written informed consent was obtained, 192 orally treated patients with type 2 diabetes without prior TZD treatment were enrolled into the trial. After randomization, they either received a fixed dose of pioglitazone (45 mg/day) in the morning or glimepiride (1 to 6 mg/day), titrated for optimal glycemic control. Inclusion criteria included an age of 40 to 75 years, HbA1c: 6.6% to 9.9%, absence of significant hepatic or renal disease, absence of congestive heart failure (New York Heart Association functional class II to IV), no cigarette smoking, and no known carotid artery disease. All study measurements were obtained at study entry and after 26 ± 2 weeks. In order to improve metabolic control, individual medical advice was given to every patient throughout the study. In the pioglitazone group, other additional oral antidiabetic therapy was permitted except for metformin, while only TZDs were excluded for additional treatment in the glimepiride group. All blood draws and measurements were performed in the morning after fasting of the patients from midnight onward.

HbA1c was determined by means of high-pressure liquid chromatography (Adams TMA1c, Menarini Diagnostics, Florence, Italy). Therapy response was defined as an absolute decrease in HbA1c from baseline by at least 0.6% (normal reference range: 4.2% to 6.0%). Glucose was assessed using a standard glucose oxidase reference method (Ruhrtal Labortechnik, Mühnesee-Delecke, Germany), and lipids (total cholesterol, low-density lipoprotein [LDL], high-density lipoprotein [HDL], triglycerides) were measured by means of standard dry chemistry methods (OSR, Olympus Diagnostica, Hamburg, Germany). Insulin was measured using a chemiluminescence method (Sciema, Mainz, Germany), PAI-I by ELISA (American Diagnostica, Pfungstadt, Germany), endothelin by ELISA (Biomedica, Vienna, Austria), and adiponectin by radioimmunoassay (Linco, St. Charles, Missouri); ELISAs from R&D Systems (Wiesbaden, Germany) were used for the determination of the following parameters: ICAM, VCAM, vascular endothelial growth factor [VEGF], MMP-9, MCP-1. The sCD40L determinations were also performed by ELISA (Bender Medsystems, Vienna, Austria). Turbimetric methods were used for the following parameters: high-sensitivity C-reactive protein (hsCRP) (Olympus, Hamburg, Germany), fibrinogen (Dade Behring, Schwalbach, Germany), and von-Willebrand factor (Instrumentation Laboratory GmbH, Kirchheim, Germany).

High-sensitive C-reactive protein changes >15 mg/l during the observation period were attributed to other inflammatory processes (flu, cold, and so on) and were eliminated before analysis.

Carotid IMT was evaluated at all time points by a single operator with high-resolution B-mode ultrasound on a single machine (Caris Plus, Esaote SpA, Genoa, Italy) with a 10-MHz linear array transducer (LA 523). All recordings were performed in a standardized way, and readings were analyzed by a physician blinded to patient profile and treatment assignment as described previously (20- 21).

The analysis of efficacy is based on the intention-to-treat population, which consists of all patients who were treated and provided assessment of the laboratory parameters at baseline and at end point of the study. All analyses were performed in an exploratory sense with appropriate parametrical and nonparametrical methods. Treatment groups were compared at baseline by using a Wilcoxon rank sum test for continuous variables and chi-square for categorical variables. Changes from baseline were evaluated by using analysis of covariance (ANCOVA) models with treatment groups as factor and baseline values as covariate. The difference between treatment groups was assessed by using t test statistics for the hypothesis that treatment group is a relevant factor in the model. Spearman rank correlation coefficients were calculated to test for the independence of the obtained results. All p values <0.05 were interpreted as statistically significant.

Of 192 patients, 179 were treated, and 173 were included into the intention-to-treat population. The most frequent reason for early termination was patient decision (seven in the pioglitazone group, three in the glimepiride group, p = 0.226). The patient characteristics of the final analysis group are given in (Table 1). There were no significant differences between the treatment groups with regard to demographic or treatment parameters. In total, 162 patients completed the protocol. Glimepiride was administered at an average daily dose of 2.7± 1.6 mg. Both treatments were well tolerated, and no episode of severe hypoglycemia was recorded. Cardiac failure requiring hospitalization was reported for two patients in the pioglitazone group.

There were no differences in all metabolic parameters at baseline. In both groups an equal and significant improvement of HbA1c was observed (pioglitazone: −0.8 ± 0.9%; glimepiride: −0.6 ± 0.8%; p < 0.001 vs. baseline in both groups, no significant difference between the groups at baseline and end point). Fasting glucose decreased at a higher extent in the pioglitazone group as compared to the glimepiride group (pioglitazone: −17.8 ± 33.5 mg/dl; glimepiride: −4.8 ± 30.8 mg/dl, p < 0.01). Fasting serum insulin concentrations decreased significantly in the pioglitazone group (−4.7 ± 6.2 μU/ml, p < 0.001), while they remained unchanged in the glimepiride treatment arm (0.4 ± 5.7 μU/ml, p = NS). A significant increase of HDL cholesterol was seen in the pioglitazone group only (from 46 ± 11 mg/dl to 54 ± 13 mg/dl, p < 0.001), while a decrease in total cholesterol was observed in the glimepiride group only (from 228 ± 38 mg/dl to 215 ± 39 mg/dl, p < 0.001). Free fatty acids, triglycerides, and adiponectin improved significantly in the pioglitazone group only. A summary of the effects of both treatment moieties on the metabolic markers is given in (Table 2).

There were no differences in all cardiovascular risk parameters at baseline. For the remaining group, a significant reduction of hsCRP was seen after six months of therapy in the pioglitazone arm (−0.10 ± 2.25 mg/dl, −29%, end point vs. baseline, p < 0.001), while no such change could be seen during glimpiride treatment (+0.24 ± 4.04 mg/dl, −4%, p = NS). Treatment with pioglitazone significantly decreased MMP-9 and MCP-1 concentrations, while glimepiride did not induce changes in these parameters. At end point, there were significant differences between the groups for both MCP-1 and MMP-9 in favor of pioglitazone (p < 0.05 in both cases). No significant changes or differences were observed for sCD40L, VEGF, ICAM, VCAM, fibrinogen, von-Willebrand factor, PAI-I, or endothelin-I. A summary of the values at baseline and end point for both treatment groups is given in (Table 3), and the percent values for the significantly changed parameters are given in (Figure 1).

While both groups were comparable at baseline, substantial regression of carotid IMT was seen in the pioglitazone treatment arm only (−54 ± 59 μm, p < 0.001 vs. baseline). The minor changes observed in the glimepiride group did not reach the level of statistical significance (−11 ± 58 μm, p < 0.001 between the groups at end point). Spearman correlation revealed a correlation between MCP-1 and carotid IMT in the pioglitazone arm (p < 0.05), while no correlation was seen between IMT reduction and hsCRP, MMP-9, or any of the other parameters. The observed changes were independent from the improvement of long-term glucose control, which is consistent with metabolic findings by other groups (22). A stratification into therapy responders (reduction in HbA1c ≥0.6%) and nonresponders (reduction in HbA1c <0.6%) revealed no difference in the overall results (Figure 2). Spearman correlation analyses between cardiovascular risk parameters and glycemic control parameters resulted in only one correlation (reduction in MMP-9 and reduction in fasting glucose control, p < 0.05) among the multiple possible combinations (Table 4).

This prospective randomized controlled trial demonstrates significant improvements of multiple cardiovascular risk markers during treatment with pioglitazone in comparison to glimepiride administration over six months. Because metabolic control, as indicated by HbA1c values, was comparably improved in both treatment arms, the observed beneficial effects of pioglitazone on cardiovascular risk markers are suggested to be independent from overall metabolic improvement but may rather be direct effects of PPARγ activation. This hypothesis is supported by the fact that the observed changes were equally seen in therapy responders and nonresponders.

In comparison to glimepiride, pioglitazone treatment led to a significant increase in HDL cholesterol and adiponectin, decrease in fasting glucose and insulin, significant higher reduction of triglycerides, and a significant higher reduction in the LDL/HDL ratio. These pronounced differences between pioglitazone and sulfonylurea treatment have been consistently described by several groups in the literature (23- 26). As one explanation for these findings, glyceroneogenesis has been recently identified as a target of TZDs in cultured adipocytes and fat tissues of Wistar rats. The activation of glyceroneogenesis by TZDs occurs mainly in visceral fat, the same fat depot that is specifically implicated in the progression of obesity to type 2 diabetes. The main role of this metabolic pathway is to allow the re-esterification of fatty acids via a futile cycle in adipocytes, thus lowering fatty acid release into the plasma (27).

Inflammation plays an important role in arteriosclerosis, and measurement of hsCRP has become a novel but emerging tool for detecting individuals at high risk for plaque rupture. A randomized placebo-controlled study with different doses of rosiglitazone resulted in a decrease of hsCRP by 26.8% (4 mg) and 21.8% (8 mg) as compared to the placebo group after 26 weeks in a Caucasian study population (28). In another recent study performed in Japanese patients, treatment with 45 mg of pioglitazone significantly reduced hsCRP by about 27% after three months of treatment, while no change occurred in the placebo comparator group. Independence from glucose metabolism was suggested by ANCOVA analysis (29). In our study, baseline values were much higher than in the Japanese study. The decrease in hsCRP by 29% during pioglitazone treatment for the first time compares directly to an equally effective antidiabetic comparator treatment that did not induce any hsCRP change.

Substantial evidence supports a causal role for MCP-1 and its receptor, CCR2, in the recruitment of monocytes from the circulation into atherosclerotic lesions. It has been shown that activation of PPAR by synthetic ligands or components of oxidized LDL reduces monocyte CCR2 expression and blocks chemotaxis mediated by MCP-1 (30). In parallel, activation of PPARγ by pioglitazone may be responsible for the observed effects in our study. While no human data has been published yet about the influence of pioglitazone on plasma MCP-1 levels, recent animal experiments have indicated and demonstrated the prevention of coronary arteriosclerosis by additional MCP-1-related anti-inflammatory effects (down-regulation of CCR2 in circulating and lesional monocytes) (16,31). In a small study without an active comparator, six weeks of treatment with 4 mg of rosiglitazone resulted in a significant improvement of plasma MCP-1 in diabetic and nondiabetic subjects (17).

While no data have been published about the influence of pioglitazone on plasma MMP-9 yet, two clinical reports describe the treatment effects of rosiglitazone on this marker. In both cases, however, the studies were only placebo-controlled, and significant differences between the treatment and comparator groups in long-term blood glucose control and HbA1c were the consequence of this design. In one study, MMP-9 decreased in a dose-dependent manner by 12.4% (4 mg dose) and 23.4% (8 mg dose) during 26 weeks of therapy as compared to placebo (28). In the other trial, 4 mg of rosiglitazone twice daily led to a significant reduction in MMP-9 by 24.1% (compared to baseline) after 12 weeks (18). In the presented study, however, the 14.5% decrease in MMP-9 under pioglitazone compares to an increase by 2.8% with a glimepiride treatment that results in the same HbA1c improvement.

No significant changes for sCD40L have been detected in our study. Reduction of this risk marker has been independently reported after treatment with rosiglitazone and troglitazone. A reduction of sCD40 by 18.4% after six weeks of treatment with rosiglitazone (4 mg twice a day) in comparison to placebo was reported by Marx et al. (19), and a mean reduction by 29% in a heterogeneous diabetic population was reported after troglitazone treatment for 12 weeks (12). However, in both study groups, the baseline values of sCD40L were about twice as high as in our study population, which may explain our inability to observe any significant differences with regard to this parameter.

No influence of pioglitazone could also be observed on PAI-1 levels. Some authors describe reduction of PAI-1 expression by TZDs (32- 33), but others found no effect on PAI-1 expression at all (34). All these observations, however, have been made in vitro. Osman et al. (35) measured PAI-1 in type 2 patients with restenosis treated with rosiglitazone and could not find any changes in their study population (35).

The major clinical finding of this study is a significant reduction of carotid IMT, a strong and well described clinical predictor of cardiovascular risk and stroke (36- 37), exclusively in the pioglitazone-treated study population. This finding and the parallel reduction in several biochemical risk markers including hsCRP, MCP-1, MMP-9, and the increase in adiponectin strongly suggest substantial antiarteriosclerotic actions of pioglitazone in vivo independent from metabolic control. While anti-inflammatory and antiarteriosclerotic effects of pioglitazone have been suggested after analysis of animal experiments (31), this is a first comprehensive clinical investigation of these effects in humans.

The answer to the question, whether the surrogate findings described in this study report can be translated into substantial clinical outcome improvements, is currently under investigation in the Prospective Pioglitazone Clinical Trial in Macrovascular Events (PROactive) study with 5,238 patients with type 2 diabetes. The cohort of patients enrolled in PROactive is a typical type 2 diabetic population at high risk of further macrovascular events. The primary end point is the time from randomization to occurrence of a new macrovascular event or death (38).

In conclusion, the presented study gives evidence of an anti-inflammatory and potential antiatherogenic effect of pioglitazone that is indicated by improvements in several traditional and nontraditional cardiovascular risk markers and carotid IMT, independent of an improvement in long-term glycemic control.