Comparison of Pioglitazone and Gliclazide in Sustaining Glycemic Control Over 2 Years in Patients With Type 2 Diabetes

This was a randomized, multicenter, double-blind, double-dummy, parallel-group, 2-year study conducted in Australia, Canada, Finland, Poland, the Slovak Republic, South Africa, and the U.K. in accordance with good clinical practice guidelines and the principles of the Declaration of Helsinki. The institutional review board at each center approved the study protocol. Patients provided written informed consent before participation.

The primary analysis was the time to failure to maintain glycemic control (HbA_1c <8%). At the time this study was designed, the American Diabetes Association (10) recommended that additional action was suggested when the patient’s HbA_1c was >8%. For this study, we chose this as the failure threshold and not the treatment target. We determined that a sample size of 300 patients per group would provide at least 80% power to detect an 11% survival difference between treatments in the distributions of time to failure to maintain glycemic control with a log-rank test (α = 0.05, two sided), assuming a 60% success rate in maintaining glycemic control in the gliclazide group. Thus, of the original 206 centers from the parent study, a subset of 98 centers, with ∼600 patients to satisfy the sample size requirements, was involved in this study (Fig. 1). The 98 centers were selected solely on the basis of the numbers of patients recruited for the parent study. Outcomes of the parent study were not known at the time of the selection of these 98 centers because the parent study was still ongoing. All patients who completed the 1-year parent study at these 98 centers were invited to participate in the extension of the parent study. Those who consented continued treatment for a further year. In the gliclazide group, 34 patients did not consent, and in the pioglitazone group, 22 patients did not consent. To avoid selection bias, all patients from the 98 centers were included in the primary analysis, regardless of whether they continued treatment for a 2nd year. This included 297 patients randomized to gliclazide and 270 patients randomized to pioglitazone at the beginning of the parent study.

The major entry criteria for the parent study were patients with type 2 diabetes inadequately controlled (HbA_1c 7.5–11.0%) with diet alone, male and female patients aged 35–75 years, and no prior use of any OAMs. There was an initial titration period of 16 weeks after the patients were randomized to taking pioglitazone or gliclazide. In the pioglitazone group, the starting daily dose was 15 mg for 4 weeks (weeks 0–4), increased to 30 mg daily for the next 4 weeks (weeks 4–8), and, finally, to 45 mg daily for the subsequent 8 weeks (weeks 8–16) on the basis of tolerability. In the gliclazide group, the starting dose was 80 mg daily for 4 weeks (weeks 0–4), 160 mg daily for the next 4 weeks (weeks 4–8), 240 mg daily for weeks 8–12, and 320 mg daily for the final 4 weeks (weeks 12–16) on the basis of tolerability. The dose of study drug was increased at each time point during titration unless the patient had not tolerated the previous dose or the investigator considered the patient at risk of experiencing hypoglycemia or other tolerability issues should the dose of study drug be further increased. In such circumstances, a reduction in dose by one titration step or maintenance of the current dose was permitted. Thereafter, the dose of the study drug was maintained throughout the remainder of the study.

We obtained fasting blood samples at each study visit for determination of HbA_1c, FPG, and FSI at a central laboratory as previously reported (11). By using HOMA, HOMA-%S and HOMA-%B were calculated (12). HOMA-%S is a surrogate for insulin sensitivity and HOMA-%B is a surrogate of β-cell activity.

The primary efficacy analysis was a comparison of the time-to-failure distributions of the two treatment groups by using a log-rank test (13). Time to failure was measured as the time from randomization to a visit at which HbA_1c was ≥8%. Failure could not occur, however, until after completion of the titration period (week 16). Thus, the earliest time to failure was 24 weeks: the first visit after the titration period.

For patients who discontinued participation from the study for reasons other than reaching the failure threshold (e.g., adverse events, withdrawal of consent), if their HbA_1c level was ≥8% before or at the time of their withdrawal, they were classified in the category of treatment failure at the time of the above threshold value. If the discontinuing patient had HbA_1c less than the failure threshold, patient data were right censored for the analysis.

Secondary efficacy analyses were performed to compare treatments regarding prespecified end points: HbA_1c, FPG, FSI, HOMA-%S, and HOMA-%B. For each of the secondary efficacy measures, a likelihood-based repeated-measures ANCOVA was conducted using the Proc Mixed implementation of SAS (version 8.2; SAS Institute, Cary, NC). The statistical model included treatment, time, treatment by time interaction, and the baseline value of the dependent variable as independent variables and employed an unstructured covariance matrix to model the within-patient error. Least-squares means were estimated for each treatment at each time point. The difference between treatments was compared at each time point based on Student’s t test (14).

The pioglitazone and gliclazide groups were similar in age, duration of diabetes, height, weight, BMI, ethnicity, and sex distribution (Table 1). Participants were mainly middle-aged, obese, white patients who had diabetes for <3 years at entry into the study. Table 1 also shows the primary reasons for discontinuation from the study in both groups. For the adverse events category, 6 of 33 (18%) patients discontinued in the pioglitazone group, compared with 1 of 25 (4%) in the gliclazide group, because of weight increase. In the gliclazide group, 3 of 25 (12%) discontinued because of headaches compared with none in the pioglitazone group.

The Kaplan-Meier curves for the two groups showed that a greater proportion of patients treated with pioglitazone maintained HbA_1c of <8% over the 2-year period than those treated with gliclazide (Fig. 2). A difference between the curves was apparent as early as week 32 and widened at each time point thereafter, becoming statistically significant from week 52 onwards. At week 104, 129 (47.8%) of 270 pioglitazone-treated patients and 110 (37.0%) of 297 gliclazide-treated patients maintained HbA_1c <8%.

Of the patients with at least one postbaseline measurement of HbA_1c, 111 of 261 (42.5%) pioglitazone-treated patients were <7% at their last visit compared with 81 of 289 (28.0%) gliclazide-treated patients (P < 0.001).

Figure 3A shows the time course of least square means for HbA_1c for both groups. Both groups had similar baseline HbA_1c. Between weeks 4 and 24, gliclazide treatment improved HbA_1c more than pioglitazone treatment did. At weeks 32 and 42, there was no difference in the HbA_1c between the two groups. From week 52 to 104, pioglitazone treatment improved HbA_1c more than gliclazide treatment did. At week 104, pioglitazone treatment lowered HbA_1c significantly more than did gliclazide treatment (pioglitazone-gliclazide HbA_1c −0.45 ± 0.11%; 95% confidence limit −0.66 to −0.23).

Figure 3B shows the time course of least square means for FPG for both groups. Both groups had similar baseline FPG. At weeks 4 and 8 (titration period), gliclazide treatment lowered FPG more than pioglitazone treatment did. Between weeks 12 and 42, there was no difference in the FPG between the two groups, except at week 16 when gliclazide treatment lowered FPG more. From week 52 to 104, pioglitazone treatment lowered FPG more than gliclazide treatment did. At week 104, pioglitazone treatment lowered FPG significantly more than gliclazide treatment did (pioglitazone-gliclazide FPG: −0.83 ± 0.22 mmol/l; 95% confidence limit −1.26 to −0.39).

Figure 3C shows the time course of least square means of body weight of patients for both groups. Both groups had similar body weight at baseline. At the end of the study, the pioglitazone group (n = 146) gained more weight than the gliclazide group (n = 127) (final weight of the two groups were 95.6 ± 0.42 vs. 93.4 ± 0.42 kg; P < 0.001).

Figure 4A shows the time course of least square means for FSI for both groups. Both groups had similar baseline FSI. In contrast to gliclazide treatment, which increased FSI, pioglitazone treatment decreased FSI from week 4 to 104. At week 104, the pioglitazone treatment group had lower FSI than the gliclazide treatment group (pioglitazone-gliclazide FSI: −52.9 ± 9.9 pmol/l; 95% confidence limit −72.6 to −33.3).

Figure 4B shows the time course of least square means for HOMA-%S for both groups. Both groups had similar baseline HOMA-%S. In contrast to gliclazide treatment, which decreased HOMA-%S, pioglitazone treatment increased HOMA-%S from week 4 to 104. At week 104, the pioglitazone treatment group had higher HOMA-%S than the gliclazide treatment group (pioglitazone-gliclazide HOMA-%S: 36.2 ± 4.4%; 95% confidence limit 27.5 to 45.0).

Figure 4C shows the time course of least square means for HOMA-%B for both groups. Both groups had similar baseline HOMA-%B. Overall, gliclazide treatment increased HOMA-%B more than pioglitazone treatment did from week 4 to 104. Between weeks 4 and 42, the difference in HOMA-%B between the two groups was substantial. After week 42, the increase in HOMA-%B in the gliclazide treatment group was much less. Hence, between weeks 52 and 104, the difference in HOMA-%B between the two groups was small. At week 104, the gliclazide treatment group had slightly higher HOMA-%B than the pioglitazone treatment group (pioglitazone-gliclazide HOMA-%B: −9.1 ± 3.7%; 95% confidence limit −16.3 to −1.82).

Several studies reported that sulfonylurea treatment, when compared with thiazolidinedione treatment, had a greater antihyperglycemic effect for as long as 4 months after therapy initiation. This effect was shown when glibenclamide (6) and glimepiride (7) were compared with pioglitazone. However, at week 52, both drugs (glibenclamide or glimepiride versus pioglitazone) showed comparable decreases in HbA_1c. This change was also observed when rosiglitazone was compared with glyburide in an open-label study (8). Glyburide resulted in an initially rapid reduction in HbA_1c, after which glycemic control deteriorated from week 16 to 52. There was no difference in HbA_1c between the rosiglitazone and glyburide groups at week 52. Similar findings were observed with gliclazide versus pioglitazone after 1 year (11). The present study extends these previous observations by analyzing time-to-treatment failure with pioglitazone and gliclazide over a 2-year treatment period and by describing the concomitant changes in HbA_1c, FPG, FSI, HOMA-%S, and HOMA-%B during this period.

These findings support the hypothesis that pioglitazone treatment is superior to gliclazide treatment in sustaining glycemic control in patients with type 2 diabetes by showing that the proportion of patients in the pioglitazone group who maintained HbA_1c <8% at any time during the 2nd year of treatment was higher than that in the gliclazide group (Fig. 2). The difference between the two treatment groups widened with each subsequent study visit. Furthermore, from week 52 onward, pioglitazone treatment improved HbA_1c and FPG more than gliclazide treatment (Fig. 3).

Kahn (15) proposed a model for the spectrum of glucose intolerance based on the relationship between insulin sensitivity and insulin secretion. As insulin sensitivity decreases, insulin secretion must proportionally increase to enable the subject to have normal glucose tolerance. If the insulin secretion is not proportionally increased, deterioration of glucose intolerance can occur. An extrapolation of this is that if insulin sensitivity is increased, insulin secretion can decrease proportionally.

HOMA is a widely used clinical tool and provides a method of assessing insulin sensitivity (or insulin resistance) and β-cell activity from basal (fasting) glucose and insulin (16). Whereas pioglitazone treatment increased HOMA-%S, it was decreased by gliclazide treatment (Fig. 3B). The difference in insulin sensitivity between the pioglitazone and gliclazide groups is sustained throughout the 2 years. In contrast, the difference in β-cell activity caused by the pioglitazone and gliclazide was not. In the gliclazide group, the initial rapid increase in HOMA-%B (which reflected the increase in β-cell activity) during the first 16 weeks of treatment (titration period) declined substantially thereafter. In the pioglitazone group, the initial increase in HOMA-%B with pioglitazone treatment was smaller compared with that of gliclazide treatment, but the increase was maintained throughout the 2-year treatment period. These changes in HOMA-%B over time resulted in a large difference between the two groups at week 16 and a smaller difference during the 2nd year (Fig. 4C). The combination of a substantial difference in HOMA-%S and a small difference in HOMA-%B from week 52 to 104 may partially explain the difference between the two treatments in glycemic control during this period.

The relative contributions of insulin-stimulated glucose uptake and basal hepatic insulin sensitivity to surrogate measures of insulin sensitivity were recently described by Tripathy et al. (17). The M value of glucose clamp studies correlated with HOMA-IR in subjects with normal glucose tolerance and in patients with type 2 diabetes. However, there was no such correlation in subjects with either impaired fasting glucose (IFG) alone or IFG/impaired glucose tolerance (IGT). The HOMA-IR correlated with hepatic insulin sensitivity in patients with type 2 diabetes and in subjects with IFG/IGT. The authors concluded that these discrepancies develop as a result of a nonparallel deterioration of the variables included in the calculation with worsening glucose tolerance.

Groop et al. (4) reported that hepatic overproduction of glucose combined with impaired peripheral glucose metabolism accounted for 43.4% of their cases. Miyazaki et al. (9) reported that pioglitazone treatment in patients with type 2 diabetes, receiving monotherapy or combination pioglitazone and sulfonylurea therapy, decreased their hepatic glucose production and increased their total body glucose disposal. Recently, Bajaj et al. (18) showed that pioglitazone monotherapy decreased hepatic glucose production and increased whole-body glucose disposal. The decrease in hepatic glucose production was negatively correlated with the increase in plasma adiponectin concentration and positively with the reduction in hepatic fat content in these patients. The increase in whole-body glucose disposal was positively correlated with the increase in plasma adiponectin concentration. Whether the enhancement of insulin sensitivity in the pioglitazone group enabled it to have a more sustained glycemic effect than did gliclazide group that improved the hyperglycemia via hyperinsulinemia remains to be established.

Other causes may contribute to the differences in sustaining glycemic improvement in patients with type 2 diabetes treated with drugs (4). Among patient-related factors, BMI and age at diagnosis may play a role. Patients with type 2 diabetes with lower BMI at diagnosis have been considered to eventually require insulin, and this has been shown for sulfonylurea “secondary failure” (19). In the present study, the BMIs of the two groups were comparable and therefore not a factor. Others have reported that the age at diagnosis may influence the duration of successful sulfonylurea therapy (20). In the present study, the mean ages of the patients were similar in both groups. Because the duration of diabetes can affect secondary failure with OAM therapy (21) or progression of type 2 diabetes, the proportion of patients who fail to maintain HbA_1c <8% at different diabetes durations may vary. In this study, the mean duration of diabetes (∼3 years) for the participants was similar in both groups. Different sulfonylureas have different secondary failure rates: gliclazide 7%, glibenclamide 17.9%, and glipizide 25.6% (22). The difference be-tween pioglitazone and other sulfonylureas in maintaining a glycemic target is not established.

In closing, results of this study show that pioglitazone is superior to gliclazide in sustaining glycemic control in patients with type 2 diabetes during the 2nd year of treatment. The combination of a substantial difference in HOMA-%S and a small difference in HOMA-%B from week 52 to 104 may explain the difference between the two treatments in glycemic control over time. Whether enhanced insulin sensitivity in the pioglitazone group promoted a more sustained glycemic effect than did gliclazide (which decreased hyperglycemia via hyperinsulinemia) remains to be established.

Takeda Europe Research and Development Centre, London, U.K., and Eli Lilly and Company, Indianapolis, Indiana, sponsored this study.

The authors thank Jane Pinaire, PhD, Neehar Gupta, MSc, Mario Widel, MS, and Todd Cravens, BA, for assistance in the preparation of this manuscript. In addition, the authors thank all of the investigators who participated in this study, without whose assistance this manuscript would not have been possible.