Pioglitazone and Rosiglitazone Have Different Effects on Serum Lipoprotein Particle Concentrations and Sizes in Patients With Type 2 Diabetes and Dyslipidemia

The study design and methods of the GLAI Study have been published (13). Briefly, inclusion criteria included patients aged ≥35 years with a diagnosis of type 2 diabetes; fasting triglyceride levels ≥150 mg/dl and <600 mg/dl; fasting LDL cholesterol levels <130 mg/dl; fasting serum C-peptide levels ≥1 ng/ml; and A1C values ≥7%, ≤11% if naïve to previous OAM therapy; or A1C values ≥7%, ≤9.5% if previously treated with OAM monotherapy. Exclusion criteria included treatment with insulin within 60 days of screening, combination OAM therapy, any lipid-altering agent, and any weight loss agent (13). This study was conducted in accordance with the Declaration of Helsinki guidelines on good clinical practice and was approved by each investigator's institutional ethical review board. All subjects gave informed consent.

Eligible patients discontinued any current OAM therapy and received oral placebo therapy throughout a 4-week, single-blind, lead-in period. Qualified personnel provided dietary counseling on the American Heart Association weight-maintaining step I diet. Randomization occurred in a stratified fashion, with four strata corresponding to previous OAM treatment (previously treated or naïve) and sex (male or female). Patients received either 30 mg PIO q.d. or 4 mg ROSI q.d. for the initial 12 weeks. For the final 12 weeks, the doses of PIO and ROSI were increased to the maximally effective doses (for monotherapy) of 45 mg q.d. or 4 mg b.i.d., respectively. In this substudy analysis, all patients who had data for baseline and end point lipoprotein subclass particle concentrations and sizes were included. The population of subjects prespecified for the analyses was all subjects randomized. We feel it is important to use this population to be consistent with the preplanned analyses and to be consistent with the intent-to-treat paradigm.

The following analyses were performed by Covance Central Laboratory Services (Indianapolis, IN): fasting blood samples (after at least 10 h of fasting) were analyzed (13) for plasma glucose, A1C, serum insulin, serum triglycerides, total cholesterol, HDL cholesterol, and LDL cholesterol. AposB and A-I (Beckman IMMAGE Immunochemistry System; Beckman Instruments, Brea, CA), Lp(a) (SPQ Antibody Reagent of Diasorin, Stillwater, MN), and apoC-III (Wako Chemicals, Richmond, VA) were determined by immunoassay. Lipoprotein subclass particle concentrations and sizes were measured using NMR spectroscopy at LipoScience (Raleigh, NC) (14,15,16). Ten subclass categories were measured in nanometers: large VLDL (>60), medium VLDL (35–60), small VLDL (27–35), intermediate-density lipoprotein (IDL) (23–27), large LDL (21.3–23.0), medium LDL (19.8–21.2), small LDL (18.0–19.8), large HDL (8.8–13), medium HDL (8.2–8.8), and small HDL (7.3–8.2). Homeostasis model assessment of insulin resistance (HOMA-IR) was determined as a surrogate of insulin resistance (17).The safety assessment has been previously reported (13).

Baseline data are presented as means ± SD. Differences between treatment groups in demographics and baseline parameters for patients entering active drug therapy were evaluated using Fisher's exact test for categorical variables or an independent-groups t test for continuous variables. For the glycemic, lipid, and HOMA-IR variables, analyses of the change from baseline level were conducted on patients for whom a baseline measurement and at least one postbaseline measurement were available. The last-observation-carried-forward change from baseline level was analyzed using a fixed-effects ANOVA, with baseline level of the analyte as a covariate (ANCOVA). The ANCOVA model comprised terms for strata, geographic region in which the investigative site was located, treatment, and baseline value. The last-observation-carried-forward changes from baseline were adjusted for their baseline level using the ANCOVA model and are reported as least-squares means with 95% CIs or with the SE of the least-squares means (18,19). Treatments were compared using the least-squares means obtained from the ANCOVA model. Spearman correlation coefficients were calculated for selected variables. SAS version 8.2 (SAS Institute, Cary, NC) was used for all analyses. All tests were two sided, and results were considered statistically significant for P < 0.05.

The patient flow through the study has been reported (13). In the PIO group 333 (of 369) patients and in the ROSI group 325 (of 366) had baseline and postbaseline lipoprotein NMR data. The distribution of patients among the various withdrawal categories was similar between treatment groups.

As in the parent study, no statistically significant differences existed between the treatment groups with respect to demographics and baseline characteristics. Data are not repeated here, as the majority of the patients from the parent study were included in this substudy.

Baseline levels; change from baseline for fasting triglycerides; total, LDL, and HDL cholesterol; free fatty acids; A1C; fasting plasma glucose; fasting serum insulin; and HOMA-IR for each group, and the comparison of changes in these parameters between groups, are similar (Table 1) to those in the parent study (13) with one exception. At baseline, all parameters were similar except for fasting triglycerides, which was significantly higher in the PIO group than in the ROSI group in this substudy. As reported previously (13), triglycerides decreased significantly in the PIO group, while it increased (not significantly in this substudy analysis) in the ROSI group. The change in triglycerides from baseline was significantly different between the two treatments. Both treatments significantly increased total and LDL cholesterol; however, PIO caused smaller increases. Both treatments significantly increased HDL cholesterol, with PIO resulting in a significantly greater increase. Compared with their respective baseline values, which did not differ between treatment groups, both treatments significantly decreased A1C, fasting plasma free fatty acids, fasting plasma glucose, fasting serum insulin, and HOMA-IR, resulting in similar end point values for both and no significant differences between treatments. Body weight changes from baseline were similar for PIO (2.0 ± 0.2 kg) and ROSI (1.6 ± 0.2 kg).

Table 2 shows the baseline, change from baseline for each group, and a comparison of these changes between them for each subfraction particle concentration.

Compared with their respective baseline values, PIO treatment significantly decreased large VLDL particle concentration but caused no change in total, medium, or small VLDL particles, whereas ROSI treatment significantly increased the total, large, and medium particle concentrations but did not change the small VLDL particles. In comparing the two treatments, the differences were significant for total, large, and medium VLDL particle concentrations but not for small VLDL particles.

PIO treatment significantly decreased IDL particle concentration from baseline, whereas ROSI treatment increased it. The between-group difference was statistically significant.

From baseline to end point, PIO treatment significantly decreased total and small, but increased large, LDL particle concentrations, whereas ROSI treatment significantly increased total and large, but caused no change in the small, LDL particle concentration. The differences between the treatments were significant for total and small LDL particles but not for large LDL particles.

Compared with their respective baseline values, PIO treatment significantly increased total, large, and medium HDL but decreased small HDL particle concentrations, whereas ROSI treatment had no effect on total HDL, significantly increased medium HDL, and decreased large and small HDL particle concentrations. The differences between the treatments were significant for total, large, medium, and small HDL particles.

Figure 1 shows the mean baseline and change from baseline for particle size of each lipoprotein class. The baseline mean VLDL particle size was larger in the PIO treatment than in the ROSI treatment group (Fig. 1A). Both treatments decreased VLDL particle sizes from baseline; however, PIO treatment decreased the mean VLDL particle size more than ROSI. The mean LDL particle sizes at baseline were comparable in both groups (Fig. 1B). PIO and ROSI treatments significantly increased mean LDL particle size; however, PIO treatment had a greater effect. The mean HDL particle sizes were comparable in both groups at baseline (Fig. 1C). There were small, but statistically significant, changes in mean HDL particle size from baseline for each group, and the difference between them was significant.

There was a negative correlation between HOMA-IR and LDL particle size in the PIO and ROSI groups and in the total study population (PIO: r = −0.249, ROSI: r = −0. 267, and combined: r = −0.252; all values P < 0.0001). There was also a negative correlation between HOMA-IR and HDL particle size in all three groups (PIO: r = −0.171, P < 0.002; ROSI: r = −0.181, P = 0.001; and combined: r = −0.165, P < 0.0001). There was a positive correlation between HOMA-IR and VLDL particle size in all three groups (PIO: r = 0.306, ROSI: r = 0.374, and combined: r = 0.339; all values P < 0.001).

PIO treatment significantly decreased apoC-III (baseline 19.3 ± 7.4 mg/dl; change from baseline −1.5 mg/dl [95% CI: −2.1 to −1.0]) and ROSI increased apoC-III (baseline 18.2 ± 6.7 mg/dl; change from baseline 1.0 mg/dl [0.4–1.6]). At baseline, PIO-treated subjects had mean apoB levels of 105.0 ± 19.9 mg/dl, with a nonsignificant change of −0.05 mg/dl (−2.4 to 2.3) during treatment. Baseline apoB was 104.0 ± 19.3 mg/dl in the ROSI-treated subjects, with a significant increase of 10.5 mg/dl (8.1–12.8) during treatment. PIO treatment had no effect on apoA-I level during treatment (baseline apoA-I: 113.8 ± 23.0 mg/dl; change from baseline 1.4 mg/dl [−0.6 to 3.4]), whereas ROSI treatment significantly decreased apoA-I (baseline 113.7 ± 22.0 mg/dl; change from baseline −5.6 mg/dl [−7.6 to −3.6]). At baseline, PIO-treated subjects had mean Lp(a) levels of 18.9 ± 22.4 mg/dl, with a significant increase of 5.6 mg/dl (4.3–6.9) during treatment. Baseline Lp(a) was 19.9 ± 23.7 mg/dl in the ROSI-treated subjects, with a significant increase of 2.8 mg/dl (1.5–4.1) during treatment. Differences between treatments for apoC-III, apoB, apoA-I, and Lp(a) were all significant.

This study is the first to demonstrate that PIO and ROSI treatments result in significantly different effects on serum lipoprotein subclass particle concentrations and sizes despite similar effects on glycemic control and insulin resistance. These observations were made utilizing NMR spectroscopy, a novel technique that permits evaluation of these important characteristics of lipoproteins, and as such has advanced our understanding of the changes in serum lipoprotein subclass particle concentration and size under normal conditions (16,20), in specific disease states (7,21–23), and during treatment with specific drug therapies (24–26).

Garvey et al. (7) reported that all three major lipoprotein subclasses differed significantly as a function of insulin sensitivity in insulin-resistant nondiabetic subjects and type 2 diabetic patients. Mean particle size of VLDL increased and the LDL and HDL particle sizes decreased as insulin resistance worsened. The dyslipidemia in type 2 diabetic patients primarily reflected an exacerbation of the changes observed in nondiabetic subjects with insulin resistance. There was a negative correlation between insulin sensitivity and VLDL particle size and a positive correlation between insulin sensitivity and LDL and HDL particle sizes. We corroborated their findings in reporting a positive correlation between HOMA-IR and VLDL particle size and a negative correlation between HOMA-IR and LDL as well as HDL particle size. However, the correlation between insulin resistance and lipoprotein particle sizes, although significant, was modest for VLDL and small for LDL and HDL.

Many studies (13) showed that PIO treatment decreases fasting triglyceride, whereas ROSI treatment increases it. One possible explanation for the difference in action of the two treatments may be their effects on VLDL particle concentration and size. Our study demonstrated that PIO treatment decreased and ROSI treatment increased large VLDL particle concentration. PIO treatment resulted in no increase in the medium VLDL particle size and concentration, whereas ROSI treatment increased them, and PIO treatment decreased mean VLDL particle size to a greater extent than ROSI treatment. Additionally, these treatments exerted differential effects on apoC-III, an inhibitor of lipoprotein lipase–mediated lipolysis and remnant clearance (27). In our study, PIO treatment resulted in a decrease in apoC-III, whereas ROSI treatment caused an increase in apoC-III, which is associated with a delay in clearance, especially of smaller VLDL and IDL particles. The greater increase in medium-sized VLDL particle concentration and IDL in the ROSI group may be partially explained by the apoC-III changes. Nagashima et al. (28) reported that PIO treatment reduced triglyceride levels, at least in part, by increasing fractional clearance of VLDL and triglyceride from circulation. They attributed the change to increased plasma lipoprotein lipase mass and decreased levels of plasma apoC-III and also suggested that PIO treatment, by increasing hepatic insulin sensitivity, could reduce expression of the apoC-III gene. A correlation between insulin resistance and apoC-III production in humans has been reported (29). That the reductions in insulin resistance caused by the two treatments were comparable in our study suggests that other mechanisms may contribute to the differences observed in triglyceride changes between PIO and ROSI treatments. Finally, it is possible that peroxisome proliferator–activated receptor α activity in PIO can account for the differences in triglyceride effects. However, there is no clear data to indicate that PIO has peroxisome proliferator–activated receptor α effects (30).

Previous studies, using methods other than NMR spectroscopy, have reported that PIO treatment (31–33) and ROSI treatment (34) increase LDL particle size. Our study, using NMR spectroscopy, is the first to directly compare the effects of these two treatments on LDL particle size and show that they had different effects not only on LDL particle size but also on concentration. The mechanism by which this differential effect on LDL particle size occurred is probably unrelated to the reduction of insulin resistance and glycemic improvement, as both drugs decreased insulin resistance and improved A1C similarly. These independent effects were demonstrated in a study in which metformin and PIO treatment resulted in a reduction in the small LDL subfraction, whereas there was no change with a sulfonylurea despite similar improvement in glycemia (33). Florez et al. (35) reported that apoC-III correlated inversely with LDL particle size, probably through effects on triglyceride-rich lipoprotein metabolism. The increase in total plasma and LDL apoB by ROSI treatment (34) suggested an increase in LDL particle concentration. We confirmed this with the NMR spectroscopy technique and extended this previous finding by showing that PIO treatment decreased LDL particle concentration, whereas ROSI treatment increased it. This may explain the smaller increase in LDL cholesterol occurring with PIO treatment than with ROSI treatment. Cromwell and Otvos (36) have reported that LDL particle concentration, as measured by NMR spectroscopy, was a strong, independent predictor of coronary heart disease and was more strongly associated with coronary heart disease risk than LDL particle size. Finally, the two treatments have divergent effects on IDL and remnants, both independent predictors of cardiovascular disease (37).

Freed et al. (34) reported that ROSI treatment increased HDL cholesterol predominantly by increasing the HDL₂ subclass, with minimal change in apoA-I level. We extend these findings by showing that 1) PIO treatment raised HDL cholesterol more than ROSI treatment; 2) PIO treatment had no effect on apoA-I, whereas ROSI treatment decreased it; 3) PIO treatment increased HDL particle size, whereas ROSI treatment decreased it; and 4) there were significant differences in their treatment effects on HDL particle concentrations. The reason for these differences in the effect between the two treatments on HDL particles, in the presence of similar reductions of insulin resistance, remains to be elucidated. It is possible that PIO treatment, in causing a decrease in triglyceride and large VLDL, reduced the cholesteryl ester transfer protein–mediated exchange of VLDL, triglycerides, and HDL cholesteryl ester, resulting in significant increases in large- and medium-sized HDL particles and a reduction in small particles, leading to a net increase in size. By contrast, the trend toward an increase in triglyceride levels among ROSI-treated patients may explain why there was a decrease in large HDL particles. The increase in HDL cholesterol in the ROSI treatment group was apparently due to a reduction of smaller-sized and an increase in medium-sized HDL particles by ROSI treatment, and a qualitatively similar change was also noted for the PIO treatment group. These changes could reflect the effect of TZDs on ATP-binding cassette-1 activity, which might be expected to produce these size changes (38). Finally, there are TZD effects that do not appear to result from an improvement of insulin sensitivity. The difference between PIO and ROSI on lipids and lipoproteins may be one of these. Whether these differences are clinically relevant is not known, since evidence for cardiovascular benefit exists for interventions that may increase larger HDL₂ particles with nicotinic acid (39) as well as those that appear to increase small HDL particles with fibrates (40).

In summary, our study demonstrates that PIO treatment and ROSI treatment differ significantly in their effects not only on serum lipids but also on lipoprotein subclass particle concentrations and particle sizes. These differences were observed despite similar improvements in many nonlipid cardiovascular disease risk factors associated with insulin resistance and type 2 diabetes. Associations of lipoprotein subclass particle concentration and size have been reported. Whether these differences in lipoprotein subclass particle concentrations and sizes translate into differences for the risk of cardiovascular disease remain to be established. Further clinical trials are needed to determine whether these differences in lipoprotein particle size results in clinically meaningful differences.

The following is a complete list of the GLAI Study Investigators: United States: Dr. Andrew J. Ahmann, Dr. Jeanine Albu, Dr. David M. Kendall, Dr. John Bernard Buse, Dr. Edward James Busick Jr, Dr. Peter J. Campbell, Dr. John V. Capotorto, Dr. John J. Cavanaugh, Dr. RoseMarie Pasmantier, Dr. A. Jay Cohen, Dr. Samuel E. Crockett, Dr. Michael H. Davidson, Dr. Mark A. Deeg, Dr. Diana G. Dills, Dr. Steven V. Edelman, Dr. Victor A. Elinoff, Dr. Keith C. Ferdinand, Dr. David Fitz-Patrick, Dr. Thomas M. Flood, Dr. Vivian Andrew Fonseca, Dr. Zachary Roy Freedman, Dr. Ronald B. Goldberg, Dr. Lisa Harris, Dr. Willa A. Hsueh, Dr. Rajeev Kumar Jain, Dr. Lois Jovanovic, Dr. Roy A. Kaplan, Dr. Robert Lang, Dr. German G. Lasala, Dr. Anne Lucia Leddy, Dr. Michael A. Lerner, Dr. Philip A. Levin, Dr. George Liu, Dr. Michael J. McCartney, Dr. Rafal Chojnacki, Dr. Edward J. Meyer, Dr. James David Mumper, Dr. Kwame Osei, Dr. David A. Price, Dr. Philip Raskin, Dr. Harvey Resnick, Dr. Wendy F. Bone, Dr. Nathaniel G. Clark, Dr. Anthony M. Roselli, Dr. Timothy R. Smith, Dr. Ralph Joe Teague, Dr. Louis H. Philipson, Dr. Joseph Ralph Tucci, Dr. Timothy Owen Wahl, Dr. Mark Lowe Warren, Dr. Jack D. Wahlen, Dr. Gary D. Yeoman, Dr. Joseph D. Krantzler, Dr. Frank N. Cole, Dr. Julio C. Pita Jr, Dr. Ayman A. Zayed, Dr. Julio Rosenstock, Dr. Victor L. Roberts, Dr. Mark S. Kipnes, Dr. Larry D. Stonesifer, Dr. Lee R.Ralph, Dr. K. Jean Lucas, Dr. Charles B. Herring, Dr. Sriranga V. Prasad, Dr. Miguel A. Franco Jr, Dr. Clifford James Houser, Dr. Arvid Lade, Dr. Richard E. Mills, Dr. Meenakshi Patel, Dr. Marina Raikhel, Dr. Rodney D. Michaels, Dr. Russell K. Havlicek, Dr. James P. Capo Jr, Dr. William Johnson Georgitis, Dr. Gregory V. Collins, Dr. John Eric Liljenquist, Dr. Kenneth G. Davis, Dr. Robert C. Hood. Puerto Rico: Dr. Elizabeth A. Barranco Santana, Dr. Vilma M. Rabell, Dr. Pedro J. Colon Ortiz, Dr. Yolanda Sierra Quinones, Dr. Miguel Sosa Padilla, Dr. Efrain Rodriguez Vigil, Dr. Shirley W. Pagan Ocasio, Dr. Jose A. Tavarez Valle, Dr. Myriam Portilla Oliva, Dr. Jorge Aguilera-Montalvo, Dr. Francisco Abreu Feshold, Dr. Luis Rodriguez Carrasquillo, Dr. Maria de Lourdes Miranda. Mexico: Dr. Maria Guadalupe Fabian San Miguel, Dr. Melchor Alpizar Salazar, Dr. Guillermo Gonzalez Galvez, Dr. Fernando Flores Lozano. Colombia: Dr. Patricio Lopez-Jaramillo, Dr. Carlos Alberto Cure, Dr. Luis Alejandro Orozco Linares, Dr. Luz Angela Casas, Dr. Monica Alexandra Terront, Dr. Diego Luis Saaibi Solano, Dr. Adriana Santamaria Cogollos.

This study was jointly funded by Eli Lilly and Company and Takeda Pharmaceuticals North America.

We acknowledge the efforts of the investigators of the GLAI Study (listed below); Lisa Martin; Sandra Althouse; Melvin Prince, MD; and Janet Tobian, MD; without whom this project could not have been completed.