Favorable Effects of Pioglitazone and Metformin Compared With Gliclazide on Lipoprotein Subfractions in Overweight Patients With Early Type 2 Diabetes

  1. James M. Lawrence, MRCP1,
  2. Julia Reid, PHD1,
  3. Gordon J. Taylor, PHD2,
  4. Chris Stirling, BSC, AIBMS1 and
  5. John P.D. Reckless, DSC, MD, FRCP1
  1. 1Diabetes and Lipid Research Group, Wolfson Centre, Royal United Hospital, Bath, U.K.
  2. 2Senior Statistician, Department of Medical Sciences, University of Bath, Bath, U.K.
  1. Address correspondence and reprint requests to Dr. J.M. Lawrence, Specialist Registrar Diabetes & Endocrinology, Southampton University Hospitals NHS Trust, Tremona Road, Southampton, U.K. E-mail: mpsjml{at}


OBJECTIVE—To compare effects of different oral hypoglycemic drugs as first-line therapy on lipoprotein subfractions in type 2 diabetes.

RESEARCH DESIGN AND METHODS—Sixty overweight type 2 diabetic patients not on lipid-lowering therapy were randomized to metformin, pioglitazone, or gliclazide after a 3-month dietary run-in. Drug doses were uptitrated for 3 months to optimize glycemia and were kept fixed for a further 3 months. LDL subfractions (LDL1, LDL2, and LDL3) were prepared by density gradient ultracentrifugation at randomization and study end. Triglycerides, cholesterol, total protein, and phospholipids were measured and mass of subfractions calculated. HDL subfractions were prepared by precipitation. The primary end point was change in proportion of LDL as LDL3.

RESULTS—HbA1c, triglycerides, glucose, and cholesterol were comparable across groups at baseline and over time. LDL3 mass and the LDL3-to-LDL ratio fell with pioglitazone (LDL3 mass 36.2 to 28.0 mg/dl, P < 0.01; LDL3-to-LDL 19.2:13.3%, P < 0.01) and metformin (42.7 to 31.5 mg/dl, P < 0.01; 21.3:16.2%, P < 0.01, respectively) with no change on gliclazide. LDL3 reductions were associated with reciprocal LDL1 increases. Changes were independent of BMI, glycemic control, and triglycerides. Total HDL cholesterol increased on pioglitazone (1.28 to 1.36 mmol/l, P = 0.02) but not gliclazide (1.39 to 1.37 mmol/l, P = NS) or metformin (1.26 to 1.18 mmol/l, P = NS), largely due to an HDL2 increase (0.3 to 0.4 mmol/l, P < 0.05). HDL3 cholesterol fell on metformin (0.9 to 0.85 mmol/l, P < 0.01). On pioglitazone and metformin, the HDL2-to-HDL3 ratio increased compared with no change on gliclazide.

CONCLUSIONS—For the same improvement in glycemic control, pioglitazone and metformin produce favorable changes in HDL and LDL subfractions compared with gliclazide in overweight type 2 diabetic patients. Such changes may be associated with reduced atherosclerosis risk and may inform the choice of initial oral hypoglycemic agent.

Patients with type 2 diabetes are frequently overweight with associated hypertension and a characteristic dyslipidemia with raised triglycerides and low HDL cholesterol. This cluster of abnormalities partly explains their increased risk of macrovascular disease. Targeting and treating the hypertension and dyslipidemia improves long-term outcomes (13). However, despite intervention, these patients remain at increased risk, suggesting that other factors contribute.

A proportion of the increased risk may be explained by qualitative changes in lipoprotein subfraction. Although LDL cholesterol may not be elevated in type 2 diabetes, the dyslipidemia is characterized by an increased proportion of LDL as small, dense LDL or LDL3 (4). Increased LDL3 has been shown to be associated with increased risk of myocardial infarction (57). Risk may also be increased by qualitative changes in HDL subfractions (8), with an increased proportion of HDL occurring as smaller, denser HDL3, which is believed to be less efficient in reverse cholesterol transport (4).

With the introduction of thiazolidinediones, there has been increased interest in the non–glucose-lowering effects of oral hypoglycemic agents (OHAs). Studies with troglitazone showed some beneficial effects on LDL subfraction distribution (911) and changes in HDL cholesterol (12). Pioglitazone treatment as monotherapy (13) or in combination therapy (14) has shown significant reductions in triglycerides and increases in HDL cholesterol (15). However, few studies have looked in detail at the effects of pioglitazone on HDL and LDL subfractions, and no comparative studies with other OHAs have been undertaken to assess whether potential improvements in lipids and lipoprotein subfractions seen with thiazolidinediones occur independent of glycemic control. A randomized, parallel-group study is reported comparing the effects of metformin, pioglitazone, and gliclazide on lipoprotein subfractions in overweight, diet-controlled type 2 diabetic patients.


A total of 67 type 2 diabetic patients aged 45–80 years were eligible and agreed to take part in the study. Eligible patients were those with diet-treated diabetes with an HbA1c >7% or those on low-dose oral hypoglycemic therapy (gliclazide up to 80 mg/day or equivalent or metformin 500 mg t.d.s.) with an HbA1c <7.5%. All patients had a BMI >27 kg/m2. Women of childbearing age had to be sterilized or using a reliable contraceptive. Patients were ineligible if 1) diet-treated with an HbA1c >10%, 2) currently taking lipid-lowering therapy, 3) previously intolerant of any study medications, or 4) study medications would be contraindicated (alanine transaminase more than three times the upper limit of normal, a serum creatinine >150 μmol/l, or a history of heart failure). Those with a recent myocardial infarction (<3 months), uncontrolled angina, or uncontrolled hypertension were excluded.

The study was a randomized, open-label, parallel-group design. Patients were seen at screening, at 6 weeks before randomization, at randomization, and at 4, 8, 12, 18, and 24 weeks after randomization. At all visits, patients were seen in the morning after at least 10 h fasting. At screening, patients were examined, and an electrocardiogram was taken. Blood pressure was measured after sitting for at least 5 min using a calibrated electronic sphygmomanometer with the average of two readings being recorded. Blood was drawn for full blood count, renal function, liver function, HbA1c, a lipid profile (total cholesterol, triglycerides, and HDL cholesterol), and glucose. All eligible patients entered a 3-month run-in phase of dietary treatment alone for their diabetes. At the visit 6 weeks before randomization, any patient who had developed symptoms suggestive of uncontrolled diabetes or who had a fasting glucose >13 mmol/l was withdrawn and commenced/recommenced on OHAs. After a total of 3 months on diet alone, blood was drawn for preparation of lipoprotein subfractions (see below) and patients were randomly assigned to either metformin 500 mg b.d., pioglitazone 30 mg o.d., or gliclazide 80 mg o.d. Patients were seen at 4-week intervals for the next 12 weeks. If fasting glucose remained >7 mmol/l, treatment was uptitrated to a maximum of metformin 1 g t.d.s., pioglitazone 45 mg o.d., or gliclazide 160 mg b.d. Patients were then followed for a further 12 weeks, when blood was retaken for lipoprotein subfractions. Patients treated with pioglitazone had liver function tests taken at each visit for safety monitoring. The study was approved by the institutional research ethics committee, and patients gave full informed written consent at screening.

Laboratory methods

Analysis of HbA1c, renal function, liver function, glucose, full blood count, and initial lipid measurements were undertaken in the central laboratories of the Royal United Hospital, Bath, U.K. The HbA1c was measured using high-performance liquid chromatography (aligned for the Diabetes Control and Complications Trial: nondiabetes range 4–6%). All other analyses were carried out in the laboratory of the Diabetes and Lipid Research Department, University of Bath, U.K. High- and low-control sera (Wako Chemicals) were run for each patient to ensure quality control. Individual three-digit patient identification numbers ensured that the laboratory staff was blinded to treatment allocation.

At randomization and at the end of the study, 20 ml of blood was drawn into EDTA for measurement of lipoprotein subfractions. Plasma was separated by centrifuging at 1,000g for 20 min at 4°C. A 1-ml sample was stored at 4°C for no longer than 4 days and used for preparation of HDL subfractions. Five milliliters were stored at −70°C for later preparation of LDL subfractions (16).

HDL subfractions

Total HDL and HDL3 were separated by a double precipitation technique first described by Gidez et al. (17). EDTA plasma (1 ml) was diluted one-to-one with 0.15 mol/l NaCl. Apolipoprotein (apo)B–containing lipoproteins were precipitated in a first step by adding heparin-manganese chloride, and HDL2 was precipitated in a second step by adding dextran sulfate. Aliquots of the supernatant at each stage were stored at −20°C for batch analysis of triglycerides, cholesterol, apoAI, and apoAII. HDL2 results were calculated.

LDL subfractions

LDL subfractions were separated by density gradient ultracentrifugation, as described by Lindgren et al. (18), using an SW40 rotor in a Beckman L8-M ultracentrifuge.

After ultracentrifugation, LDL subfractions (LDL1, LDL2, and LDL3) were stored at −20°C for later measurement of triglycerides, cholesterol, apoB, total protein (pyrogallol red; Randox Laboratories; interassay coefficient of variation 4.2%), free cholesterol, and phospholipids. The total mass of each fraction was calculated by adding triglycerides, cholesterol, total protein, and phospholipids. Total cholesterol, apoB, and triglycerides were also measured in serum at randomization and at the end of the study.

Statistical analysis

Data were analyzed using SPSS version 11. All normal data are expressed as means ± SD. The distribution of BMI (weight in kilograms divided by height in meters squared) was not normally distributed, and results are expressed as median (interquartile range). Triglycerides were log transformed before analysis. The primary analyses on LDL and HDL subfractions were undertaken using ANCOVA for between-group comparisons, applying a Bonferroni correction to allow for multiple groups. As BMI was not normally distributed, ANCOVA of the ranked data were used to compare between groups for BMI at baseline and over time.

A secondary analysis for within-group comparisons was made using Student’s t test. The secondary analysis has not been adjusted for multiple comparisons. These results need to be considered with care and require further study because the trial was not adequately powered to fully look at these effects. For all analyses, the level of statistical significance was set at P < 0.05 (two tailed).

Sample size calculation

As very little data are available for the effects of OHAs on lipoprotein subfractions, sample size calculations were based on anticipated changes in triglycerides, which have been shown to be related to changes particularly in LDL subfractions. In the U.K. Prospective Diabetes Study, the average triglyceride concentration in patients with diet-controlled diabetes was 1.8 mmol/l (19). From previous studies using pioglitazone monotherapy (20), we estimated a fall in triglycerides from mean 1.8 to mean 1.5 mmol/l with improved diabetes control. We expected that the effect of metformin and gliclazide would be smaller, with an estimated fall in triglycerides from mean 1.8 to mean 1.7 mmol/l (21). It was anticipated that the range of the change in triglycerides would be between 0 and 1 mmol/l in the three patient groups. From this, a conservative estimate of the SD of the change in triglycerides would be 0.16 mmol/l. Assuming this to be the case and using a Bonferroni correction to allow for the multiple group comparisons, 14 patients would be needed in each group to have 80% power at the 5% significance level to detect the anticipated difference between the groups. We aimed to recruit and randomize 20 patients in each treatment arm to allow for dropouts. Patients already in the dietary run-in phase once 20 patients were randomized to each drug treatment were continued in the study.


Sixty-seven patients were eligible and entered the dietary run-in phase. Three withdrew at 6 weeks in the dietary run-in phase due to hyperglycemia (fasting glucose >13 mmol/l). Two in the gliclazide group withdrew after randomization; one with symptomatic hyperglycemia requiring addition of a second oral agent at 12 weeks and one with a myocardial infarction after 4 weeks who was started on insulin. One patient in the metformin group died of an acute myocardial infarction at 6 weeks postrandomization, and one patient in the pioglitazone group withdrew as a result of increasing ankle edema 4 weeks after randomization. In all, 20 patients in each drug group completed the study. In the gliclazide group, eight patients remained on the starting dose of 80 mg o.d. throughout the study, six were uptitrated to 160 mg/day, one was uptitrated to 240 mg/day, and five were uptitrated to 320 mg/day. In the pioglitazone group, seven remained on the starting dose of 30 mg throughout the study and 13 were uptitrated to 45 mg. In the metformin group, two remained on 1 g/day, three were uptitrated to 1.5 g/day, four were titrated to 2 g/day, and 11 were uptitrated to 3 g/day. The rest of the results relate to those who completed the study.

Table 1 shows the baseline demographic details. The age and sex distributions were comparable across groups. The same proportions in all groups had treated hypertension or a blood pressure >140/80 mmHg, the threshold at which treatment would be considered if confirmed on repeat measurement, although a higher proportion of patients in the metformin and gliclazide groups were on antihypertensive drug treatment at baseline.

Baseline measures of HbA1c, BMI, cholesterol, triglycerides, and HDL were comparable (Table 2). A significant fall in HbA1c was seen in all groups, with no significant difference in the mean change in HbA1c over time. The BMI fell significantly in the metformin group and there was a significant rise with pioglitazone and gliclazide. Although there was a fall in triglycerides in both the metformin and pioglitazone groups, this did not reach statistical significance, and the change was not different from that in the gliclazide group. Total cholesterol fell significantly on metformin, but again, this change was not different from that seen in the other groups. Total HDL cholesterol increased significantly in the pioglitazone group. Comparing the three groups, the change seen with pioglitazone was significantly different from that seen with metformin and gliclazide (P = 0.001 for between group comparison), largely due to the difference between pioglitazone and metformin (P = 0.026).

Table 3 shows the changes in subfractions. No change in total LDL cholesterol was seen in any group. LDL3 mass and the proportion of LDL as LDL3 fell significantly over 6 months in both the pioglitazone and metformin groups, and these changes were significantly different from those of the gliclazide group (P = 0.001, pioglitazone versus gliclazide; P = 0.001, metformin versus gliclazide; P = NS, metformin versus pioglitazone). In the pioglitazone group, the reduction in LDL3 was associated with a reciprocal increase in LDL1 mass and proportion (P = 0.03, pioglitazone versus gliclazide; P = NS, gliclazide versus metformin; P = NS, metformin versus pioglitazone). Changes in LDL subfractions were independent of changes in HbA1c, triglycerides, and BMI.

The increase in HDL cholesterol on pioglitazone was largely due to an increase in HDL2 cholesterol (Table 3). There were no significant changes in this group in apoAI and AII, but there was a significant reduction in the ratio of apoAI to apoAII in both total HDL and HDL3. Comparing the three drug groups, the change in total HDL apoAI/AII with pioglitazone was significantly different from that seen with metformin and gliclazide (P = 0.035; pioglitazone versus metformin, P = 0.05; pioglitazone versus gliclazide, P = NS). Although total HDL cholesterol remained unchanged on metformin, there was a significant reduction in HDL3 cholesterol and a nonsignificant increase in HDL2 cholesterol. In both the pioglitazone and metformin groups, the HDL2-to-HDL3 ratio increased from baseline compared with that of the gliclazide group.


We have shown significant and potentially important changes in lipoprotein subfraction distribution in overweight patients with type 2 diabetes when treated for 6 months with pioglitazone and metformin but not with gliclazide.

Unlike previous studies, we have not shown a significant fall in triglycerides with pioglitazone. Although total cholesterol was reduced in the metformin group, the changes seen were not significantly different from those in the pioglitazone and gliclazide groups, and there was no difference between groups in LDL cholesterol. Nevertheless, for the same degree of improvement in glycemic control, we have demonstrated significant changes in the LDL subfraction profile in patients treated with pioglitazone and metformin, independent of changes in BMI or triglycerides. Our study is the first to use density gradient ultracentrifugation to compare LDL subfraction profiles across different OHA groups, with the advantage over other methods of separating LDL fractions of being able to quantify changes in mass. To date there has been surprisingly little data assessing the effects of older OHAs on LDL subfractions. There are no previous data with gliclazide or with metformin as monotherapy. The only data so far with metformin are from a study (22) where metformin was added second line in patients inadequately controlled on glyburide alone. Gradient gel electrophoresis was used to assess changes in peak particle diameter as a marker of changes in LDL subfractions, and the results for metformin were neutral. Although these results cannot be compared directly with ours because patient groups and methods of assessing subfraction profiles were different, further evidence will be needed before firm conclusions can be drawn about the overall effects of metformin on LDL subfractions. It is of interest that although we saw a significant fall in LDL3 mass and LDL3 proportion on metformin, the total cholesterol–to–apoB ratio, which can be used as a surrogate marker for changes in LDL subfraction distribution, remained unchanged in this group. This contrasts with the significant increase in the cholesterol-to-apoB ratio with pioglitazone, suggesting an increase in size of LDL particles and therefore a shift from LDL3 toward larger, less dense (and potentially less atherogenic) LDL. Previous data are also largely lacking on the effects of pioglitazone on LDL subfractions, although the changes we have seen in our study are in line with those using troglitazone in patients with type 2 diabetes (10), as well as obese patients without diabetes (9) and those with insulin resistance after a myocardial infarction (11).

It would be anticipated that the reduction in LDL3 on pioglitazone and metformin would be associated with reduced atherosclerosis risk. Increased LDL3 is associated with increased rates of myocardial infarction (5,6,7). LDL3 is more prone to oxidation and glycemic modification (23) and as a result does not bind as well to hepatic LDL receptors (24) and is more likely to bind to the extracellular matrix and be taken up by scavenger macrophages in atherosclerotic plaques (2527).

The three drug groups differed in their effects on total HDL and HDL subfractions. HDL cholesterol is recognized as an important predictor of macrovascular disease, with increased risk being associated with a fall in total HDL cholesterol and also an increase in the total cholesterol–to–HDL ratio. Outcome studies (28,29) have shown that directly targeting and increasing HDL can improve long-term outcomes, both in those with isolated low HDL and in those with mixed dyslipidemia with high cholesterol, low HDL, and high triglycerides. Consistent with other studies, we have seen a significant increase in total HDL with pioglitazone. This is in contrast to the neutral effect with metformin and gliclazide, and such a change would be expected to be antiatherogenic. Although the change in the total cholesterol–to–HDL ratio was of borderline significance on pioglitazone, this change was not different from that seen on either metformin or gliclazide.

HDL particles are a heterogeneous group that can be separated by density (HDL2 [1.063–1.125 g/ml] and HDL3 [1.125–1.21 g/l]) or by apo content (apoA1 without AII or apoA1 with AII). An increase in HDL2, the HDL2-to-HDL3 ratio, and apoAI seem to be particularly associated with a reduction in macrovascular risk (8,30). Although the total HDL cholesterol did not change in the metformin group, the HDL3 cholesterol fell significantly and in association with this, the HDL2-to-HDL3 ratio significantly increased. Similarly, the increased total HDL on pioglitazone was associated with a significant increase in HDL2 cholesterol, a nonsignificant fall in HDL3, and a significant rise in the HDL2-to-HDL3 ratio. The changes in HDL subfractions with pioglitazone are consistent with the increased activity of lipoprotein lipase seen in a number of studies using thiazolidinediones (31,32), and similar changes have been described with rosiglitazone monotherapy (33). Data on enzyme activity changes are more limited with metformin but there is some evidence, particularly from animal studies, of similar increases in lipoprotein lipase activity with metformin to those seen with thiazolidinediones, which may explain the increased HDL2-to-HDL3 ratio seen on metformin (34).

There was no change in either total apoAI or apoAII in any group over time. ApoAI and apoAII both fell in the metformin group in the HDL3 subfraction, although the apoAI-to-apoAII ratio remained the same. The significance of this finding, along with the reduction in apoAI-to-apoAII ratio in total HDL and HDL3 on pioglitazone and in HDL3 on gliclazide is uncertain because epidemiological and outcome data so far only relate to changes in apoAI and AII in total HDL and there are no data looking at differences in apoAI and apo AII in HDL subfractions or apoAI-to-apoAII ratios.

Our study is the first to compare the effects of the three commonly used oral hypoglycemic drug classes on lipoprotein subfractions in early type 2 diabetes in a randomized parallel-group design, and we have shown potentially beneficial effects of both pioglitazone and metformin on HDL and LDL subfractions compared with gliclazide. In view of the available outcome data, metformin clearly remains the drug of first choice in overweight patients with type 2 diabetes. Where metformin is not tolerated or contraindicated, our data would support use of a thiazolidinedione, and particularly pioglitazone, ahead of a sulfonylurea, as it is possible to achieve similar glycemic control with additional benefits on macrovascular risk factors.

Many patients with diabetes are treated with oral combination therapy and also with lipid-lowering therapy. It would be informative to study the effects of different drug classes in combination on lipoprotein subfractions and to assess whether the effects of metformin and pioglitazone seen in our study would be additive to those of lipid-lowering therapy, potentially enhancing the reduction in macrovascular disease.

Table 1—

Baseline demographic data of patients completing the study

Table 2—

Changes in routine laboratory and examination data over the course of the study

Table 3—

Lipoprotein subfraction changes


This study was designed by the investigators, who acknowledge a grant-in-aid to the department from Takeda Pharmaceuticals.


  • J.P.D.R. has been on advisory panels for and received honoraria from Takeda Pharmaceuticals.

    A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.

    • Accepted October 10, 2003.
    • Received June 17, 2003.


| Table of Contents