Effect of Rosiglitazone Versus Insulin on the Pancreatic β-Cell Function of Subjects With Type 2 Diabetes

After approval of the study by the Institutional Review Board of the University of Alabama at Birmingham and appropriate written informed consent, we randomized 17 subjects with type 2 diabetes, who were inadequately controlled on a regimen of a sulfonylurea and metformin, to the addition of an insulin injection of 70/30 mixed human insulin administered once daily before supper or 8 mg of rosiglitazone administered once daily. The dose of rosiglitazone was fixed for the remainder of the study, whereas the 70/30 insulin was started at 0.2 units/kg and adjusted to achieve a fasting blood glucose level of ≤120 mg/dl without occurrence of severe or frequent hypoglycemia. All subjects remained on their randomized therapy for 6 months with adjustment in dose only occurring in the insulin group to correct for hypoglycemia. When titration of the insulin dose was completed, all subjects remained at this dose for 6 months unless significant and/or frequent hypoglycemia occurred when the dose was reduced and fixed at this level for the remainder of the study unless further hypoglycemia occurred.

Fasting glucose, serum insulin, and proinsulin levels were measured at baseline and at the end of the study. Plasma glucose level was measured by a glucose oxidase method using a YSI glucose analyzer (Yellow Springs Instruments, Yellow Springs, OH). Serum insulin was assayed using a competitive chemiluminescent immunoassay (DPC, Los Angeles, CA). Proinsulin was measured by enzyme-linked immunosorbent immunoassay (Nichols Institute, San Clemente, CA). At baseline and after 6 months of therapy, a intravenous glucose tolerance test (fsIVGTT) was performed to determine the acute insulin response to glucose (AIRg) as a way to evaluate first-phase insulin response to glucose. Additionally, a glucagon stimulation test for C-peptide was performed 1 week after the fsIVGTT. C-peptide was measured using a double-antibody C-peptide kit (reference range 1.0–5.0 ng/ml). All tests were performed in the fasting state, and the subjects in the insulin group were asked to withhold their insulin injection the night before testing.

During the fsIVGTT, venous blood samples for measurement of serum insulin were collected at baseline (average of −10 and −4 min samples) and at +2, +3, +4, +5, +6, +8, and +10 min after an intravenous bolus of glucose calculated at a dose of 300 mg/kg of body weight given over 60 s starting at time 0. The AIRg was calculated as the area under the curve above the baseline (AUCab) during the first 10 min after the glucose infusion.

During the glucagon stimulation test, venous blood samples for the measurement of C-peptide were collected at baseline and 6 min after an intramuscular injection of glucagon.

Insulin resistance was calculated at baseline and at 6 months using the following homeostasis model assessment (HOMA) of insulin resistance (HOMA-IR) formula: (insulin [micro–International Units/milliliter] × glucose [millimoles/liter])/22.5. For calculation purposes, the reciprocal of IR (1/IR) was used as the insulin sensitivity index (S_i).

β-Cell function was determined by the disposition index, which was calculated as the product of AIRg and S_i. Statistical analysis was performed using two-tailed Student’s t tests for most comparisons except for the fsIVGTT, which was analyzed using a one-way ANOVA and performed using GraphPad Prism Version 4.00 for Windows (GraphPad Software, San Diego, CA). Data are shown as means ± SE. P values <0.05 were considered statistically significant.

As shown in Table 1, the groups were well matched for age, BMI, and duration of diabetes. Furthermore, and most importantly, HbA_1c was well matched among study groups before the start of the study and decreased to the same level after treatment in both groups (Table 2, Fig. 1).

The AIRg increased significantly, from a baseline of 1.42 to 16.71 μIU · ml⁻¹ · 10 min⁻¹ at 6 months (+15.3 μIU · ml⁻¹ · 10 min⁻¹ [AUCab]; P ≤ 0.001), in the rosiglitazone group but not in the insulin group, in which there was a nonsignificant decrease from a baseline of 8.43 to 7.23 μIU · ml⁻¹ · 10 min⁻¹ at 6 months (−1.2 μIU · ml⁻¹ · 10 min⁻¹ [AUBab]; NS) (Fig. 2).

As expected, rosiglitazone induced a significant improvement (92.3% increase) in insulin sensitivity (Table 3). S_i was calculated using the reciprocal of the HOMA method as described above.

Furthermore, the disposition index increased significantly in the rosiglitazone-treated group from a baseline of 0.18 to 4.18 at 6 months (P = 0.02); meanwhile, the insulin-treated group experienced a nonsignificant decrease from a baseline of 1.86 to 1.23 at 6 months (NS) (Fig. 3).

Although no differences between groups were observed when looking at the raw C-peptide data obtained during the glucagon-stimulated C-peptide tests, we found significant differences when adjusting for the differences in insulin sensitivity and calculating the disposition index. The rosiglitazone-treated group demonstrated a significant (+43%) increase in the disposition index compared with a nonsignificant decrease (−9.4%) in the insulin group after 6 months of therapy (Table 4).

Furthermore, the proinsulin-to-insulin ratio, which was equally elevated in both groups at baseline, decreased significantly in the rosiglitazone group after treatment (Fig. 4).

We have therefore demonstrated, in a prospective randomized and controlled study, that rosiglitazone improves pancreatic β-cell function after 6 months of therapy and that this improvement is independent of glycemic control because with a similar reduction in HbA_1c no improvement in β-cell function was found in the insulin-treated group.

Dysfunctional pancreatic β-cells are known to release more proinsulin; therefore, the ratio of proinsulin to insulin generally increases when β-cell function decreases. Therefore, the decrease in the proinsulin-to-insulin ratio observed in this study may be interpreted as evidence of improved β-cell function confirming previous studies in which the HOMA-β improved and the proinsulin-to-insulin ratio decreased with rosiglitazone but remained unchanged or increased with placebo, metformin, or sulfonylurea (3,4).

A defect in the first phase of insulin release is usually the first measurable sign of β-cell dysfunction that can be observed in both type 1 and type 2 diabetes. A return or improvement of the first-phase insulin secretory response to glucose has not been previously demonstrated with other antihyperglycemic agents including secretagogues, metformin, or α-glucosidase inhibitors. Although others have reported that therapy with multiple daily injections of insulin can lead to improvement in the proinsulin-to-insulin ratio similar to that we observed with rosiglitazone, these improvements have been short-lived (9 weeks) and failed to demonstrate improvements by other methods (5).

Alternatively, other investigators have reported no improvement in insulin secretion after treatment with rosiglitazone; however, this study was probably too short in duration to demonstrate an effect on β-cell function. Furthermore, the small number of subjects might have caused a type 1 error because the investigators reported an improvement on insulin sensitivity and a tendency to an increase in the disposition index in the rosiglitazone-treated group (6).

Therefore, this is the first study to demonstrate an improvement in the first phase of the insulin secretory response to glucose in subjects with long-standing type 2 diabetes receiving oral pharmacological therapy. This study confirms our previous observational finding in a case-control study that the addition of a TZD, but not the addition of metformin, improves glycemic control as well as endogenous insulin production (1). This is consistent with our findings in a continuing long-term observational study demonstrating that the addition of a TZD to a failing oral regimen of sulfonylurea and metformin results in significant improvements in glycemic control in most subjects with type 2 diabetes. Furthermore, this improvement is maintained for a prolonged period of time, as long as 5 years, which we hypothesize is due to improved endogenous insulin production (2).

We believe that the mechanism responsible for the improvement in endogenous insulin production is directly related to the lowering of serum and tissue free fatty acid (FFA) levels that is known to occur with TZDs but not with other antihyperglycemic agents including insulin secretagogues, α-glucosidase inhibitors, and biguanides. Animal studies have shown that β-cell failure is preceded by an increase in plasma FFAs and accumulation of triglyceride in β-cells (7). In addition, if lipid accumulation in the β-cell is reversed by creating a hyperleptinemic state, β-cell function improves and development of diabetes is avoided (8). The damage to the β-cell is caused by elevated β-cell levels of FFA, its metabolite ceramide, nitric oxide, and peroxinitrite levels, leading to accelerated β-cell apoptosis (9,10). In addition, increased FFA levels decrease the expression of the IDX-1 gene, which is responsible for the formation of new β-cells from stem cells in the pancreatic duct (11). Rosiglitazone has been shown in animal studies to decrease islet cell triglyceride levels and increase stainable insulin in both the db/db mouse and the Zucker diabetic fatty rat models (12,13).

Autopsy studies of human β-cells have shown that with aging, more fat accumulates in the pancreatic β-cells of the islets of Langherhans than in the α-cells or the pancreatic duct cells and that apoptosis rather than decreased formation of β-cells is responsible for the decreasing β-cell mass seen in type 2 diabetes (14,15). Therefore, in the absence of unethical pancreatic biopsy studies in type 2 diabetic patients, we must assume that the mechanism of pancreatic β-cell destruction and its correction with rosiglitazone is similar to that seen in animal models.

Because rosiglitazone and perhaps other TZDs differ from the currently available therapies for type 2 diabetes in their ability to rejuvenate pancreatic β-cells, these drugs should be used at the earliest possible opportunity in the course of type 2 diabetes and not withheld until the later stages of the disease (16). In this study, the average duration of diabetes was 7.6 years. It should be noted that determining the exact date of onset of type 2 diabetes may be a difficult task and, in general, type 2 diabetes goes unrecognized for several years before being diagnosed. In this study, the time of onset was determined by patient recollection and review of medical records.

There are potential limitations of this study that need to be discussed. One of these is the difference (9 years) in the mean age between the two treatment groups; although this difference was not statistically different, a negative effect of aging on β-cell function has been shown by some investigators (17). Nonetheless, the fact that the baseline AIRg was better in the insulin group and that both groups had similar proinsulin-to-insulin ratios at baseline seems to make this age difference irrelevant. Another potential limitation of this study is the lack of determination of hepatic insulin extraction, which can potentially account for differences in proinsulin-to-insulin ratio (18).

In conclusion, this study demonstrates that rosiglitazone, but not insulin, helps to induce a recovery of pancreatic β-cell function, as evidenced by the restoration of the first-phase insulin response to glucose. This effect was in independent of the correction of glucose toxicity.

This work was supported by a research grant from GlaxoSmithKline Pharmaceuticals.