The Combined Effect of Triple Therapy With Rosiglitazone, Metformin, and Insulin Aspart in Type 2 Diabetic Patients

Sixteen obese type 2 diabetic outpatients treated with human NPH or MIX insulin twice daily were recruited from our university diabetes clinic for this investigator-initiated study (Table 1). The study was approved by the local ethical committee and by the Danish Medicines Agency. All subjects gave written informed consent. The inclusion criteria were age 40–76 years, fasting serum C-peptide >250 pmol/l, BMI 25–40 kg/m², HbA_1c >7%, and insulin therapy, whereas the exclusion criteria were s-creatinine >100 μmol/l, plasma alanin aminotransferase >100 units/l, intolerance to metformin, left ventricular ejection fraction measured with echocardiography <40%, blood pressure >180/100 mmHg, severe dyslipidemia (total cholesterol >8 and triglyceride >5 mmol/l), and unawareness of hypoglycemia. A reference material of age- and BMI-matched nondiabetic control subjects (11) was used in comparison with the degree of insulin resistance (Table 1).

At inclusion, subjects were matched for HbA_1c and age and were thereafter block randomized into two groups of eight subjects, the triple therapy group or to continued NPH or MIX insulin therapy (control group) (Table 1). During a 2-month start-up period, when the pretests were carried out, the treatment was unchanged in both groups. A total of seven patients already received metformin before randomization, three in the control group and four in the triple therapy group.

Following randomization, the control group continued their insulin regimen (and the three control group patients using metformin before randomization also continued this treatment), but the insulin dose was increased to obtain the best possible metabolic control. The aims for the control group were blood glucose within the range of 5–7 mmol/l preprandial and HbA_1c <7% without inducing severe hypoglycemia (i.e., events needing help from another third person). One to two mild hypoglycemic events were accepted per week. Self-monitored blood glucose (SMBG) was measured four times per day, before meals and at bedtime.

The triple therapy group started with rosiglitazone (Avandia) 4 mg/day (dose was increased to 8 mg/day after 8 weeks), metformin (Glucophage) 0.5 g twice per day (dose was increased after 4 weeks to 1 g twice daily), and insulin aspart (NovoRapid) three times per day injected just before each main meal. Initially, total daily insulin dose was reduced to one-third of the dose given during the start-up period to avoid hypoglycemia during the start of treatment with rosiglitazone and metformin, and it was divided into three equal doses. In the triple therapy group, the target SMBG values were 5–7 mmol/l postprandial (measured 1.5 h after each meal) and HbA_1c <7%, with no severe hypoglycemic events but one to two mild events were accepted per week.

The insulin dose in both groups was adjusted by the diabetologist at scheduled visits and at scheduled telephone contact (five outpatient clinic visits and two telephone contacts) and kept constant in the last 2 months of the study period. Thus it was possible to change insulin dosages seven times during the study. Overall, insulin dose was adjusted as follows: a reduction in total insulin dose of ∼10% if mean blood glucose was <5 mmol/l, unchanged if blood glucose was 5–7 mmol/l, and increases in insulin dose of ∼5–10% if blood glucose was 7–10 mmol/l, 10–20% if blood glucose was 10–13 mmol/l, and 20–30% if blood glucose was >13 mmol/l. In the control group, the morning insulin dose was changed mainly based on prelunch, predinner, and nighttime SMBG values, whereas the late insulin dose was changed mainly based on fasting SMBG. In the triple therapy group, the insulin aspart dose before a specific meal was determined by the postprandial SMBG values at that meal.

Subjects were asked not to change their diet and exercise level during the study period. HbA_1c was measured every second month, and a 24-h profile of blood glucose, serum insulin, serum C-peptide, and plasma free fatty acid (FFA) was carried out before and after 6 months of treatment. Insulin sensitivity was measured before and after the 6-month treatment period.

This test was performed as previously described by Hother-Nielsen et al. (15). A primed-constant intravenous infusion of 3-³H-glucose (DuPont-New England Nuclear, Boston, MA) was started in the morning and continued throughout the entire study. The ratio of priming dose (10 ml) and constant infusion (0.1 ml/min) was 100 (15). To achieve a common level of basal plasma–specific activity, the tracer infusion was adjusted for surface area by adjustment of the infusate volume (16). After a 120-min basal tracer equilibration period, insulin (Actrapid Human; Novo Nordisk) was infused at a rate of 40 mU · min⁻¹ per m² body surface area for 240 min. After initiation of the insulin infusion, the plasma glucose concentration was allowed to fall to 5.5 mmol/l, at which level it was maintained (clamped) using a variable infusion of 20% glucose enriched with 3-³H-glucose to maintain the specific activity constant at baseline levels during the clamp (17).

Continuous indirect calorimetry, using a ventilated hood system (Deltatrac 2; SensorMedics, Yorba Linda, CA) was performed during the last 40 min of the basal and insulin infusion periods. Urine samples for determination of the urea excretion were obtained from the start to the end of the euglycemic-hyperinsulinemic clamp for calculation of protein oxidation rates.

Beginning at 8:00 a.m., one catheter was inserted into an antecubital vein for blood sampling of blood glucose, serum insulin, serum C-peptide, and plasma FFA. Blood samples were obtained right before each main meal (breakfast, lunch, and dinner) and then 30 min, 1, 1.5, 2, 2.5, and 3 h later and thereafter every hour until the next meal. At night, blood samples were obtained at 3:00 and 7:00 a.m.

Plasma glucose concentration was measured using the glucose oxidation method (Glucose Analyzer 2; Beckman Instruments). HbA_1c was measured by a high-performance liquid chromatography method (Tosho A1c 2.2; Tosho Bioscience, Japan) (normal range 4.4–6.4%). Serum insulin and C-peptide samples were analyzed by a two-site, time-resolved immunofluorometric assay (DELFIA) (18). Insulin aspart was measured with a modification of the method described by Andersen et al. (19). Plasma FFA concentrations were measured by Itaya et al.’s method (20). Plasma total cholesterol, HDL cholesterol, and triglyceride concentrations were measured by a kit from Boehringer Mannheim (Diagnostica, Mannheim, Germany), and LDL cholesterol was calculated using the Friedewald equation. Tritiated glucose-specific activity was determined on barium/zinc deproteinized plasma samples as previously described.

Rates of glucose appearance (R_a) and rates of glucose disappearance (R_d) were calculated using Steele’s non–steady-state equations, using a distribution volume of 200 ml/kg (21) and a pool fraction of 0.65 (22). Glucose storage rate was calculated as R_d minus the glucose oxidation rate. The insulin sensitivity index (S_i) was calculated as R_d divided by incremental increase in serum insulin concentration during the insulin clamp and divided by the mean glucose concentration during the last hour of the insulin clamp. Hepatic glucose production (HGP) was calculated as: HGP = R_a − glucose infusion rate.

Results are presented as means ± SE unless otherwise stated. Differences between the groups or within a group were compared by unpaired or paired T tests. Correlation analysis was performed using Pearson correlation analysis. Statistical analysis was performed using SPSS for Windows (version 10.0). P values <0.05 were considered significant.

HbA_1c did not change in the control group during the study period, whereas in the triple therapy group, a marked reduction was observed (8.8 ± 0.9 to 6.8 ± 0.3%, P = 0.004) (Fig. 1). Thus, at the end of the study period, HbA_1c levels in the triple therapy group were close to the reference value (4.4–6.4%) and were 2.0% points better than in the control group on twice daily insulin treatment (P < 0.001). Twenty-four-hour blood glucose profiles were sampled before and after 6 months of intervention (Fig. 2A). Before intervention, the two curves were superimposable (data not shown), but after intervention, blood glucose in the triple therapy group was <6 mmol/l during most of the 24-h period, and the postprandial peaks of hyperglycemia were generally eliminated. During the night, glucose values increased slightly in both groups and to a slightly higher level in the triple therapy group. However, the fasting blood glucose concentrations in the morning in the two groups after 6 months of intervention were identical (7.8 ± 0.9 and 8.2 ± 0.8 mmol/l, respectively).

In the control group, the insulin dose was increased gradually and reached the level of 0.67 ± 0.24 units · kg body wt⁻¹ · day⁻¹ (at inclusion 0.46 ± 0.16 units · kg body wt⁻¹ · day⁻¹, P < 0.03), whereas the insulin dose following intervention was unchanged in the triple therapy group (0.29 ± 0.11 vs. 0.42 ± 0.16 units · kg body wt⁻¹ · day⁻¹) despite improvement in metabolic control and being significantly lower than the insulin dose in the control group (P < 0.002).

The diurnal profiles of serum insulin were superimposable in the two groups before intervention (data not shown). After 6 months, the pattern for the control group was unchanged, but the diurnal level was increased (Fig. 2B). However, after 6 months, the triple therapy group showed clearly defined insulin peaks at meal times. The peaks rose rapidly up to 400 pmol/l after 60 min and thereafter rapidly reduced again in the pattern quite similar to that seen in nondiabetic subjects (Fig. 2B). The insulin peaks were mainly due to the insulin aspart injections (Fig. 2C). These measurements also showed that insulin aspart given with the evening meal was cleared again before midnight. Therefore, plasma insulin concentrations during the night were much lower in the triple therapy group than in the control group. This difference may be a safeguard against nocturnal hypoglycemia.

After 6 months of intervention, glucose infusion rates in the triple therapy group rose significantly from 101.8 ± 22.0 to 166.5 ± 11.8 mg · min⁻¹ · m⁻², (P < 0.02); however, insulin sensitivity was still below the level for the nondiabetic reference group (∼270 ± 20 mg · min⁻¹ · m⁻²). When insulin sensitivity was estimated by the “tracer technique,” S_i for the triple therapy group was found to increase too, and this increase was mainly due to an increase in glucose oxidation and not in glucose storage (Table 2). In accordance with this, lipid oxidation during the insulin clamp decreased significantly (Table 2). Interestingly, lipid oxidation during the clamp correlated with mean diurnal FFA concentration in the combined group of patients (R = 0.65, P < 0.01).

During intervention in both groups, HGP was suppressed to the normal level (nondiabetic control group: 21 ± 3 mg · min⁻¹ · m⁻²) but at lower serum insulin values in the triple therapy group than in the control group, indicating improved liver insulin sensitivity during treatment in the triple therapy group (Table 2).

Body weight increased nonsignificantly by 1 kg in the control group and 3 kg in the triple therapy group (P > 0.1). Blood pressure and lipids did not change significantly (Table 2). Fasting plasma FFA values were identical before and after intervention in the triple therapy group. However, during the 24-h profile mean diurnal plasma FFA values were lower in the triple therapy group compared with the control group (22.7 ± 1.4 vs. 33.5 ± 4.7 mmol/l, P = 0.057).

A slight decrease in the hemoglobin concentration in the triple therapy group was found (Table 2), and interestingly, plasma alanin aminotransferase values decreased (but not statistically significantly), which may be caused by a reduction of fat content in the liver.

One subject in the triple therapy group stopped treatment due to subjective sensation of fluid retention but no peripheral edema was demonstrated and X-ray of the chest was without sign of pulmonal edema.

Despite the increased insulin dose in the control group and the improvement of metabolic control in the triple therapy group, no severe hypoglycemic attacks were reported in either group. In the last 2 months, when the insulin dose was maintained constant and glucose levels were stable, one mild hypoglycemic event was reported in the control group and a total of seven events (in two subjects only) in the triple therapy group (P > 0.1). No nocturnal episodes were registered.

In this study of poorly controlled, long-term type 2 diabetic subjects with severe insulin resistance and abolished first-phase insulin response, we found that a new concept of treatment with insulin aspart, metformin, and rosiglitazone based on pathophysiological knowledge was able to significantly improve the HbA_1c level without inducing severe hypoglycemia. In fact, the regimen may be a safeguard against nocturnal hypoglycemia. No changes in body weight, blood pressure, or lipid profile were observed. The only side effect recognized was a 7% reduction of the hemoglobin concentration. This new concept shows interesting potential, but of course long-term studies are needed to prove this (these studies have already been initiated in Denmark).

Our primary aim was to test the hypothesis that normalization of the three major defects in the pathophysiology of type 2 diabetes may result in near normalization of blood glucose. We found that the HbA_1c levels in the triple therapy group after 6 months improved significantly. Furthermore, the 24-h glucose values measured at home (SMBG) and during a 24-h profile in the hospital (Fig. 2A) showed a glucose pattern close to that found in normal subjects with geometric mean values of about 7 and 6 mmol/l, respectively. Thus, our study seems to confirm the hypothesis that restoration of the three pathophysiologic defects in type 2 diabetic subjects may improve the blood glucose profile.

Insulin sensitivity was not completely normalized in the triple therapy group compared with a nondiabetic population, but insulin aspart at mealtime was able to mimic a normal insulin profile with rapid and high insulin peaks. This resulted in no postprandial glucose elevation after breakfast and lunch and minor elevation after dinner. These findings demonstrate the importance of a normal insulin profile but also show that improvement of postprandial glucose values has an impact on the 24-h glucose profile as well as on fasting blood glucose. Moreover it shows the putative impact of postprandial hyperglycemia on HbA_1c.

With respect to nocturnal hypoglycemia, triple therapy seems to be safe because the risk is very low. The reason for this is insulin aspart, which disappears from the blood during the evening and therefore only endogenous insulin is present at night. If blood glucose declines at night, endogenous insulin secretion declines as well.

In the control group, HbA_1c values did not improve significantly despite a 50% increase in NPH or MIX insulin dose. The explanation for this is primarily that these subjects have severe insulin resistance; some received 100 units/day of insulin. The reason for the high HbA_1c values despite high doses of NPH insulin could be that postprandial hyperglycemia is not avoided by 1–2 injections of NPH insulin per day, as indicated in a recent study by Yki Järvinen et al. (23). Thus, NPH insulin once or twice daily seems unable to normalize blood glucose values. It could be argued that the dose of insulin in the control group was still not high enough to produce a normal glycemic response. However, our study demonstrates that with triple therapy postprandial glucose and HbA_1c decreased without an increase in insulin dose, which is completely opposite to what is seen in the control group. Therefore, the answer to our secondary aim is that triple therapy is associated with a lower HbA_1c level than treatment with NPH or MIX insulin, at least in the present study setup.

In the literature, we found 25 studies on combination therapy in type 2 diabetes, i.e., a combination of NPH insulin and either metformin, sulphonylureas, acarbose, or glitazone running for time periods of 3–12 months. Overall, after treatment in these studies, HbA_1c levels only decreased to a value of ∼9% (6). Thus, the results of the combination of NPH insulin and one oral antidiabetic drug have not been encouraging. The combination of NPH insulin at night with metformin twice daily seems the most appropriate treatment for the moment, but even in the best of these studies, the HbA_1c levels are 1.2% above the upper level for control subjects (6).

Only two studies on the effect of rapid-acting insulin before meals have been carried out in type 2 diabetic subjects. In a controlled study, the combination of regular insulin at meals and a sulphonylurea was not better than NPH insulin alone (24). A similar result was obtained with the combination of insulin lispro at meals with a sulphonylurea in comparison with the combination of NPH insulin with a sulphonylurea (25). Altogether, the literature and our results suggest that triple therapy seems superior to other regimes, yet it has to be tested in larger studies against different combinations of treatment.

Type 2 diabetic subjects are insulin resistant, i.e., insulin-mediated glucose disposal is significantly reduced, as also demonstrated in this study (10). This defect seems to be allocated to skeletal muscle, where glucose uptake, disposal, and oxidation are reduced. In this study, glucose oxidation during treatment in the control group was unchanged, whereas glucose oxidation improved significantly in the triple therapy group (Table 2). This makes sense because an increase in glucose oxidation is mandatory for a decline in blood glucose values. The increase in glucose oxidation may be mainly explained by rosiglitazone treatment (26). In accordance with this, FFA levels during the day were reduced during treatment with triple therapy, and the mean FFA concentration correlated with lipid oxidation during the clamp. This may be mainly explained by the treatment with rosiglitazone.

Basal HGP has been reported to be slightly increased (by about 15%) in untreated type 2 diabetic subjects (11,27). In this study, fasting HGP was slightly increased compared with nondiabetic reference material. After 6 months, fasting HGP was still within the normal range despite a decline in fasting serum insulin in the triple therapy group, indicating that other variables also play an important role in fasting HGP. The action of HGP during the insulin clamp was estimated at only two insulin concentrations in this study, but the results indicate that the liver is also insulin resistant and that the treatment increases hepatic insulin sensitivity in the triple therapy group. The improvement of insulin sensitivity in the liver may be related mainly to metformin treatment (26).

Body weight increased slightly but nonsignificantly. Improved metabolic control can by itself account for an increase in body weight. Yki Järvinen (6) calculated that a 1% decrease in HbA_1c level results in a 2-kg increase in body weight. Thus, there is no reason to anticipate that rosiglitazone should have increased body weight specifically. However, if so, this effect may have been counterbalanced by metformin treatment.

Blood pressure and plasma triglyceride did not change significantly, and there was no difference between the groups. Specifically, there was no increase in LDL cholesterol in either group. Rosiglitazone did not induce any degree of liver toxicity whatsoever. One subject stopped the triple therapy after several months due to a subjective feeling of edema, but objective signs of edema were not found and X-ray of the chest was normal. The only recognized side effect was the well-known reduction in hemoglobin concentration.

The main aim in this study was to test a hypothesis, and this does not involve a controlled blinded protocol. However, a subaim was to study if the chosen concept, i.e., triple therapy, was better than treatment with NPH or MIX insulin twice daily. Such a study should ideally be blinded, but that is impossible if you choose to give both groups the benefit of changing insulin dosages during the study period. Furthermore, in a blinded trial, syringes with saline should have been used, from which ethical considerations arise. By offering both groups identical care, applying identical goals for the metabolic control in both groups, and using insulin algorithms, we feel that we have compensated for the missing blinding, and therefore a bias in favor of the triple therapy group has been minimized. One year after completion of the study, five subjects still receive triple therapy, and these subjects’ glycemic control is nearly identical to what it was at the end of the study (6.7 ± 0.4 vs. 6.6 ± 0.3%), whereas seven subjects in the control group have continued conventional treatment with no improvement in HbA_1c over time (8.8 ± 0.4 vs. 9.2 ± 0.8%)

In conclusion, we have shown that 6 months of treatment with a new concept based on our available pathophysiological knowledge (rosiglitazone, metformin, and insulin aspart) significantly reduced the 24-h glucose profile and the HbA_1c level. No significant increase in body weight and no impairment in plasma lipids, blood pressure, and safety parameters (hemoglobin was an exception) were observed in the triple therapy group.

These results strongly indicate that normalization of the three major pathophysiological defects of type 2 diabetic subjects, namely peripheral insulin resistance, hepatic insulin resistance, and reduced insulin response following meals can significantly improve glucose metabolism. Triple therapy seems a promising treatment for hyperglycemia in type 2 diabetic subjects, but long-term studies are necessary. However, if these results are confirmed, diabetes complications may be dramatically reduced.

Insulin aspart was kindly supplied by Novo Nordisk, Denmark, rosiglitazone by GlaxoSmithKline, U.K., and metformin by Merck, U.K. This study was supported by grants from the Clinical Institute, University of Southern Denmark, Bernard Rasmussen and wife Meta Rasmussen’s foundation, Overlæger˚odets Legatudvalg, Den Almindelige Danske Lægeforening’s science foundation, Novo Nordisk, Denmark, and GlaxoSmithKline, Denmark.

The authors gratefully acknowledge Klaus Levin, Kurt Højlund, Tony Bill Hansen, Jørgen Hangaard, Knud Erik Pedersen, Karin Dyregaard, Charlotte Olsen, Lone Hansen, and Eva Lund for their valuable contributions. A special thanks to Ole Blaabjerg for developing the insulin aspart assay.