Exploring the Substitution of Exenatide for Insulin in Patients With Type 2 Diabetes Treated With Insulin in Combination With Oral Antidiabetes Agents

This exploratory, multicenter, two-arm, parallel-design, open-label trial was conducted over 16 weeks at five centers in the U.S. Patients randomized to the exenatide group used a multi-use pen to subcutaneously inject a fixed dose of 5 μg b.i.d. for 4 weeks and 10 μg b.i.d. for the remaining 12 weeks of the study (before morning and evening meals). Patients in the reference group remained on their insulin regimens throughout the 16-week study. No specific glycemic goals were set for insulin patients during the trial. Patients in both treatment arms continued their oral antidiabetes medications and were instructed to continue their current diet and exercise regimens. As per protocol instructions, the sulfonylurea dose was decreased by ∼50% in response to one confirmed or two suspected (symptoms without confirmatory blood glucose <3.4 mmol/l) hypoglycemic events.

Eligible patients were between 30 and 75 years of age, had been diagnosed with type 2 diabetes for ≥2 years, and had been treated with one of the following for ≥3 months to 12 years: once- or twice-daily NPH insulin, once-daily insulin glargine, once- or twice-daily ultralente insulin, or an insulin mixture. All patients were on oral antidiabetes regimens consisting of an immediate- or extended-release metformin and/or a sulfonylurea for at least 3 months before screening or a fixed-dose sulfonylurea/metformin combination therapy. Additional inclusion criteria included, at the time of screening, an A1C level ≤10.5%, BMI >27 and <40 kg/m², and a history of stable body weight. Patients were excluded if they had more than three episodes of severe hypoglycemia within 6 months before screening, had used any prescription drug to promote weight loss within 3 months, or had previously received exenatide or glucagon-like peptide 1 analogs.

A common clinical protocol was approved by an institutional review board at each site and was conducted in accordance with the principles described in the Declaration of Helsinki (18). Patients were recruited according to local practices, and all participants gave written informed consent before participation. A1C levels were measured at weeks −2, 0 (randomization and study initiation), and 16 (or at early discontinuation). Blood chemistries and fasting serum lipids were assessed at weeks −2 and 16 (or at early discontinuation). Fasting serum glucose and fasting C-peptide were assessed at weeks −2, 2, 4, and 8. Weight, vital signs, concomitant medications, and drug doses were collected at week 0 and at each subsequent visit. Patients performed 5-point self-monitored blood glucose (SMBG) profiles at study weeks 2, 4, 8, and 16.

Adverse events were assessed at each visit and were reported as preferred terms from the Medical Dictionary for Regulatory Activities. A clinical trial adverse event was defined as any untoward medical occurrence, without regard to the possibility of a causal relationship. A hypoglycemic episode was defined as any time a patient felt that he or she was experiencing a sign or symptom of hypoglycemia or noted a blood glucose level <3.4 mmol/l (60 mg/dl) during self-monitoring, regardless of whether this level was associated with signs, symptoms, or treatment.

The primary hypothesis was that >60% of the patients who stopped insulin therapy and initiated exenatide twice daily would maintain glycemic control. It was estimated that 30 patients, randomized in a 2:1 ratio (exenatide to insulin reference), would be sufficient to verify the probability of observing >60% success in the exenatide group. The function of the reference group was to provide additional confidence in the validity of changes in A1C observed in those using exenatide. SAS (version 8.2; SAS Institute, Cary, NC) was used to conduct all statistical analyses. Tests of within-group change (last observation carried forward) were based on the Wilcoxon's signed-rank test (with a two-sided significance level of 0.05). Student's t tests (with a two-sided significance level of 0.05) and Fisher's exact tests were used for between-group comparisons. We conducted a post hoc analysis of hypoglycemia based on events confirmed by glucose measurement. The overall exposure rate (total confirmed events divided by total exposure) was calculated using two criteria: 1) the American Diabetes Association criteria (confirmed blood glucose <3.9 mmol/l) and 2) the a priori study design criteria (confirmed blood glucose <3.4 mmol/l).

The intention-to-treat sample included all randomized patients with type 2 diabetes who received at least one dose of the study drug. Patients were considered to have maintained glycemic control if they did not experience a clinically relevant rise in A1C (an increase of ≥0.5%) at end point. Patients were also considered to have maintained glycemic control if they maintained their A1C levels below the prespecified limit but discontinued from the study after week 8 for reasons other than loss of glucose control. Patients were considered to have lost glycemic control if they experienced a clinically relevant rise in A1C or if they discontinued from the study at any time because of loss of glycemic control (as determined by the investigator). A stepwise logistic regression procedure was used to identify variables associated with maintenance of glycemic control in the exenatide group.

A total of 51 patients with type 2 diabetes were randomized (2:1) to exenatide or insulin reference therapy (online appendix Fig. 1 [available at http://dx.doi.org/10.2337/dc06-2532]). A total of 45 patients (29 in the exenatide and 16 in the insulin reference groups) had sufficient data for the primary efficacy analysis; 4 discontinued before their true glycemic control outcome could be determined. Table 1 presents baseline demographics, clinical characteristics, and background therapy. Exenatide-treated patients who successfully maintained glycemic control (n = 18) had (on average) shorter disease duration, higher pretreatment fasting C-peptide levels, were taking comparatively less insulin, and were observed to have been receiving insulin therapy for a shorter period of time.

The overall mean A1C change in the exenatide group (n = 29) was +0.3 ± 1.5% (within-group change, P = NS). The overall mean A1C change in the insulin group (n = 16) was −0.1 ± 0.7% (within-group change, P = NS). The mean changes in A1C were not significantly different between groups at end point. Of the patients who substituted exenatide for insulin, >60% (18 of 29, mean A1C change −0.5 ± 0.7%, P = 0.003) successfully maintained glycemic control as defined a priori, supporting the primary hypothesis of the study. Of the 16 patients in the insulin reference group, 13 (81%) maintained glycemic control. Individual changes in A1C and fasting serum glucose are shown in online appendix Fig. 2. The majority of exenatide-treated patients who maintained control also completed the 16-week study (14 of 18) and had reductions in A1C from baseline (12 of 18). Four patients were observed to have successfully maintained glycemic control at week 8 but discontinued before week 16 for reasons other than loss of glycemic control (nausea [one patient] or patient decision [three patients]).

Of the 11 exenatide-treated patients who did not maintain glycemic control (mean A1C change +1.6 ± 1.5%, P = 0.001), 5 discontinued the study before week 16 because of loss of glycemic control (as determined by the investigator) and 6 lost glycemic control as evidenced by exceeding the predefined A1C criterion. Within-group analyses demonstrated that the subset of exenatide-treated patients unable to maintain glycemic control had a significant increase in fasting glucose by week 2 (+3.9 ± 2.7 mmol/l, P = 0.005), which increased through week 8 (+5.2 ± 3.3 mmol/l, P = 0.008). However, neither an end point rise in fasting glucose nor an earlier rise in week 2 fasting glucose was significantly associated with treatment failure in logistic regression analyses. Of the characteristics described in Table 1, pretreatment C-peptide (maximum likelihood estimate 2.96, P = 0.024) and baseline body weight (maximum likelihood estimate −0.07, P = 0.088) were identified as the best predictors of successful glycemic control.

Most of the exenatide-treated patients (27 of 29, 93%) lost weight during the study. Less than one-half of the patients (6 of 16, 38%) in the insulin reference group lost weight. Exenatide patients experienced a steady decline (Fig. 1) in mean body weight (end-point change −4.2 ± 3.0 kg; within-group change, P < 0.001), while the mean body weight with insulin was not substantially changed (end-point change +0.5 ± 1.7 kg, P = NS). The between-group comparison of mean changes in body weight at end point was statistically significant (P < 0.001). The mean changes in body weight observed in the two exenatide groups (those who maintained vs. those who lost glycemic control) were similar (between-group change, P = NS). We conducted a post hoc analysis to determine the degree of correlation between change in A1C and change in body weight. The Pearson correlation coefficients were −0.29 and −0.25 for the exenatide and insulin reference groups, respectively.

As observed in the SMBG profiles (Table 2), patients in the insulin reference group demonstrated consistently better prebreakfast blood glucose control (P < 0.05) at all study follow-up visits (weeks 2, 4, 8, and 16) compared with that in exenatide-treated patients. Conversely, exenatide-treated patients demonstrated better postprandial glucose control after breakfast (P < 0.05) at study weeks 2, 8, and 16, with less consistent improvement demonstrated following the evening meals (i.e., better postprandial control compared with insulin only at week 8, P < 0.05).

In the exenatide group, 26 of 33 (79%) patients reported a treatment-emergent adverse event compared with 9 of 16 (56%) in the insulin reference group. Most of the adverse events reported in this study were mild to moderate in intensity. Adverse events considered possibly related to exenatide treatment were predominantly gastrointestinal in nature (e.g., nausea, diarrhea, vomiting, abdominal pain, gastroesophageal reflux disease), with nausea the most common (48.5% incidence). Five exenatide-treated patients discontinued from the study because of an adverse event (nausea [three patients], bronchitis, and hyperglycemia). Two serious adverse events (chest pain and excessive hyperglycemia) were reported in the study, both occurring in the same exenatide-treated patient; this patient required hospitalization and discontinued because of hyperglycemia before completing the week-16 study visit. There were no adverse events considered possibly related to insulin treatment. Headache was the most common adverse event (31.3% incidence) reported in the insulin reference group, followed by nausea (12.5%) and cough (12.5%)

The incidence of hypoglycemia was 39% (13 of 33) and 38% (6 of 16) in the exenatide and insulin reference groups, respectively. Most of the hypoglycemia was reported to have occurred during the daytime (exenatide, 11 of 13 patients; insulin, 4 of 6 patients). Of the 13 exenatide-treated patients who reported hypoglycemia, 10 were taking concomitant sulfonylurea. A total of 12 hypoglycemia events (8 in the exenatide and 4 in the insulin groups) were confirmed by glucose measurement (<3.4 mmol/l). The overall hypoglycemia rates were 1.72 and 0.97 events/patient-year for the exenatide and insulin reference groups, respectively. The rates were slightly higher in the subgroups that maintained glycemic control (exenatide, 2.54 events/patient-year; insulin, 1.18 events/patient-year). There were no episodes of serious hypoglycemia (i.e., requiring medical intervention). One patient treated with exenatide and sulfonylurea had three severe hypoglycemic episodes (i.e., episodes that required assistance of another person and were associated with either a blood glucose level <50 mg/dl or prompt recovery after oral carbohydrate, intravenous glucose, or glucagon injection), which were treated with food or drink; this patient discontinued after week eight because of nausea.

In this exploratory study, substitution of exenatide for insulin therapy resulted in no deterioration in glycemic control in ∼62% of the type 2 diabetic patients studied. The remaining 38% of patients did experience deterioration in glycemic control. In some, this deterioration manifested as a rise in A1C noted at study end (after 16 weeks), and in others the worsening of hyperglycemia occurred by 2–8 weeks after insulin withdrawal. In those patients where there was no deterioration in glycemic control, a potential benefit for carrying out such substitution of therapy was the observed weight loss. Although A1C was not increased at end point in the majority (15 of 29, 52%) of the exenatide-treated patients, A1C remained above target (>7.0%) in many (20 of 29, 69%). Most of the insulin-treated patients (9 of 16, 56%) also did not experience an increase in A1C; however, all but 1 patient (15 of 16, 94%) remained above target (A1C > 7%) at end point. The number of exenatide-treated patients who both improved glycemic control and lost weight (11 of 29) through this therapeutic substitution was relatively small. The adverse event profile in patients who switched to exenatide was consistent with the predominance of gastrointestinal side effects observed in prior phase III, placebo-controlled trials. The overall incidence of hypoglycemia was similar between the exenatide and insulin reference groups, and most of the patients in the exenatide group who experienced hypoglycemia were taking concomitant sulfonylurea. It is known from the phase III trials studies of exenatide that hypoglycemia occurs, most commonly, when exenatide is used with sulfonylurea.

Several factors should be considered when interpreting our results. First, the glycemic control results may have been better if we had added exenatide to the treatment of patients on insulin and decreased insulin doses gradually rather than replacing insulin with exenatide abruptly. This choice may have also contributed to the relatively high dropout rate in the exenatide group. Secondly, this study was not designed to compare insulin treatment with exenatide; thus, there were no specific glycemic goals set for insulin patients during the trial. The small number of patients in the trial also makes drawing conclusions from between-group statistical comparisons problematic. Finally, given the small sample size and exploratory nature of this study, it was not possible to adequately answer all clinically relevant questions related to this type of therapy substitution. For example, since only five patients treated with exenatide had a baseline A1C <7%, it would be difficult to conclusively determine whether it is possible to sustain glycemic control below the target level.

It should be noted that while stable glucose control was observed in a majority of patients, deterioration in glycemic control was observed in some patients, and many patients remained above target (A1C >7.0%). Exenatide elicits its glucose-lowering effect in part by glucose-dependent stimulation of insulin release from pancreatic β-cells (11,19). It is not clear whether glucose-dependent stimulation of insulin is the dominant glucose-lowering mechanism compared with suppression of glucagon or slowing of gastric emptying, for example, but it has been observed to be an important action in the hierarchy of possible mechanisms. Given that type 2 diabetes is a progressive disease where β-cell function gradually diminishes over time (1), it is possible that exenatide may have a diminished capacity to exert a glucose-lowering effect in patients with more advanced disease and minimal β-cell function. For example, patients who require insulinization during the night may be more likely to experience deterioration in glycemic control if cessation of insulin is attempted. It is therefore interesting to note that exenatide-treated patients less likely to have a favorable outcome in the current study had (on average) longer disease duration, lower pretreatment fasting C-peptide levels, were taking comparatively more insulin, and were observed to have been receiving insulin therapy for a longer period of time. However, it is not possible to characterize clear predictors of outcome given the size and exploratory nature of the study.

The weight effect observed in this study is also of interest. Prior clinical trials have shown that exenatide treatment results in weight loss as a monotherapy and when added to a variety of background oral therapies including metformin, sulfonylureas, and thiazolidinediones (20–23). Greater weight loss is seen in exenatide-treated individuals with metformin background therapy. Data from 2-year open-label extension trials have demonstrated that the weight reduction associated with exenatide is progressive over time and is associated with improvements in a number of cardiovascular risk factors (24). In the current study, early (week 2) weight reduction was observed in the patients who maintained glycemic control after coming off insulin, suggesting that the weight effect may be associated with both exenatide treatment and insulin withdrawal. The current study demonstrates that a substantial number of patients treated with insulin will experience deterioration in glucose control when exenatide is substituted. However, the study does not ascertain the mechanism by which weight loss occurs.

In conclusion, this exploratory study gives some insight into the outcome of substituting one injectable therapy (exenatide) for another (insulin) in patients with long-standing type 2 diabetes. The majority of patients maintained glycemic control, although most did not fully optimize control. Several patients experienced deterioration in glycemic control, indicating that therapy substitution is not feasible for all patients currently treated with insulin. In this study, patients who transitioned from insulin to exenatide needed to perform SMBG and were given parameters as to when to contact their physician. Although the exploratory nature of this study limits its predictive power, the results suggest that patients with longer disease duration who are taking higher doses of insulin and who have less endogenous β-cell reserves are less likely to have a favorable outcome with this therapy substitution. Further investigation is warranted.

We wish to thank the patients, investigators, and their staffs for participating in the study.