β-Cell Function in Mild Type 2 Diabetic Patients: Effects of 6-month glucose lowering with nateglinide

Subjects in the present study were a subset of those who had participated in a trial of the effects of nateglinide in mildly hyperglycemic patients with type 2 diabetes (9) and who had undergone a mixed meal study as part of that protocol. With 32 patients originally assigned per group, 108 patients completed the substudy (30 in the placebo group and 26, 27, and 25 each in the 30-, 60-, and 120-mg dose groups, respectively). Only select sites participated in the meal challenge substudy, and treatment assignment was randomized within a site. To minimize the possibility of a systematic bias, patients were studied and data were analyzed concurrently. Men and women aged >30 years with a medical history of type 2 diabetes who were maintained on diet alone for at least 6 weeks before screening were recruited. Patients were selected to have a BMI between 22 and 35 kg/m² and agreed to maintain their prior diet and exercise habits during the entire study. Patients were included in the double-blind active treatment period if their mean fasting plasma glucose concentration was between 7.0 and 8.3 mmol/l upon screening and during the 4-week placebo run-in. Patients were excluded if they had a history of type 1 diabetes, diabetes that resulted from pancreatic injury, or acute metabolic or significant diabetes complications. Patients were also excluded if they had received oral antidiabetic treatment during the 3 months or chronic insulin treatment during the 6 months before screening and had known sensitivity to drugs similar to nateglinide or had medical conditions that precluded their participation in the trial (9). Informed consent was obtained from all participants, and the study was performed in accordance with the Declaration of Helsinki and the rules governing medicinal products in the European Community following institutional review board approvals.

The study was a randomized, placebo-controlled, double-blind, parallel-group design trial. Patients were randomized in approximately equal numbers at week 0 to receive 30, 60, or 120 mg nateglinide or placebo taken before breakfast, lunch, and dinner for 24 weeks. This was preceded by a 4-week, single-blind, run-in period during which all patients took placebo before the three main meals. On the study day at weeks 0 and 24, predose blood samples were drawn following an overnight fast, and then the study medication was given. Ten minutes later, patients were given a standard 475-kcal breakfast (9). Blood samples were obtained at 15, 30, 60, 90, 120, and 240 min after breakfast for the determination of plasma glucose, insulin, and C-peptide concentrations, as described previously (9).

Glucose tolerance was assessed as the fasting plasma glucose concentration, the mean postprandial glucose concentration during the 4 h of the meal test, and the incremental postprandial glucose response (the difference between the two). An empirical parameter of glucose-induced insulin release, namely the incremental insulin-to-glucose concentration ratio at 30 min postglucose (or insulinogenic index ΔI₃₀/ΔG₃₀), was also calculated. Areas under glucose, insulin concentration, or secretion rate curves were calculated by the trapezoidal rule.

The β-cell model used in the present study, describing the relationship between insulin secretion and glucose concentration, has been illustrated in detail previously (5,6). Insulin secretion, S(t), consists of two components: S(t) = S_g(t) + S_d(t).

The first component, S_g(t), represents the dependence of insulin secretion on absolute glucose concentration (G) at any time point and is characterized by a dose-response function, f(G), relating the two variables. Characteristic parameters of the dose response are insulin secretion at a fixed glucose concentration of 7 mmol/l (approximately the fasting glucose level in our mildly diabetic subjects) and the mean slope in the observed glucose range, denoted here as glucose sensitivity. The dose response is modulated by a potentiation factor, P(t), that accounts for several potentiating factors (prolonged exposure to hyperglycemia, nonglucose substrates, gastrointestinal hormones, neurotransmitters, and effects of nateglinide itself). The first secretion component is thus described by the following equation: S_g(t) = P(t)f(G).

The potentiation factor is set to be a positive function of time and to average one during the experiment. It thus expresses a relative potentiation of the secretory response to glucose. In previous experiments (11), we have found that insulin secretion is relatively higher at the end of a meal or glucose test than at the beginning, when compared with the glucose levels (i.e., when glucose returns to the basal level insulin secretion remains higher). Potentiation thus increases during the test; this increase has been quantified as the ratio of potentiation values at two time points during the meal (5,6). In this study, as we also found an early drug-induced increment in the potentiation factor, we used the ratio of the potentiation factor value at 30 min to that at 0 min. This index quantifies the potentiation of the early insulin release rather than the traditional potentiation over the entire test (5,6).

The second insulin secretion component represents a dynamic dependence of insulin secretion on the rate of change of glucose concentration expressed as:

This component is termed derivative component and is determined by a single parameter (k_d), denoted as rate sensitivity. Rate sensitivity also is related to early insulin release (5,6).

The model parameters [the parameters of the dose response f(G) and the potentiation factor P(t)] were estimated from glucose and C-peptide concentrations by regularized least squares, as previously described (5,6). Regularization involves the choice of smoothing factors that were selected to obtain glucose and C-peptide model residuals with SDs close to the expected measurement error (∼1% for glucose and ∼4% for C-peptide). As the treatment code was known when the present analysis was done, to limit possible bias the smoothing factors were chosen in each subject not knowing the relationship between the current subject’s parameter values and those previously calculated. Thus, the determination of posttreatment parameters was not influenced by the knowledge of the results pretreatment. As we have previously shown (5), this parameter estimation procedure resulted in reasonable reproducibility of parameter estimates. Coefficients of variation were 16% for insulin secretion at fixed glucose concentration, 24% for glucose sensitivity, and 52% for rate sensitivity. From the estimated model parameters, total and basal insulin secretion was calculated. Insulin secretion was expressed in picomoles per minute per meters squared of body surface area.

Insulin sensitivity was indirectly estimated using homeostasis model assessment (12) as the product of fasting glucose and insulin concentrations, and from glucose and insulin levels during the meal with the use of oral glucose insulin sensitivity (OGIS; equation for a 2-h oral glucose test with a glucose dose of 75 g) (13). The validity of OGIS for a meal test was previously verified in a group of 43 normal subjects with an eightfold span in insulin sensitivity as assessed by the euglycemic insulin clamp technique. In this group, OGIS was well correlated with the M value from the clamp (r = 0.59, P < 0.0001) (A.M., O. Schmitz, and E.F., unpublished data).

Data are given as means ± SD. Due to their skewed distribution, insulin parameters are expressed as median (interquartile range). Differences in baseline clinical and metabolic characteristics were tested by using the χ² test for nominal variables and the Kruskall Wallis test for continuous variables. The effects of the four treatments were compared by ANCOVA, which included effects for treatment, baseline measure, and treatment-by-baseline interaction. When the latter term was not statistically significant, the analysis was rerun without it. Preliminary homoscedascicity tests were run to control for unequal group variances. In post hoc testing, individual between-group differences were tested by contrasts (with P value adjustment for multiple comparisons). The contribution of treatment-induced changes in parameters of β-cell function to corresponding changes in plasma glucose concentrations was assessed by stepwise multiple regression analysis of the whole dataset.

Baseline clinical and metabolic characteristics did not differ significantly across treatment groups (Table 1). Over 4 h following meal ingestion, mean postprandial plasma glucose concentrations (8.9 ± 1.9 mmol/l, n = 108) were 1.4 ± 1.5 mmol/l above the fasting levels, an average increment of 18 ± 2%. Fasting plasma insulin concentrations (which were 102 [75] pmol/l) rose to an average postprandial value of 276 (211) pmol/l during the meal, an approximate twofold increase.

In placebo-treated patients, glucose tolerance, as both the fasting and postprandial glucose concentration, deteriorated modestly over the 24 weeks of study (Table 2). Nateglinide produced a dose-related reduction of mean postprandial glucose levels; in addition, active treatment was associated with decreases in fasting glycemia that reached statistical significance only at the highest dose (Table 2). In contrast, plasma insulin concentrations, whether fasting, postprandial, or incremental, showed no significant change in any of the treatment groups (Table 2). The insulinogenic index (ΔI₃₀/ΔG₃₀) showed a trend toward an increase, which fell short of statistical significance (Table 2). The time course of plasma insulin concentrations during the meal test at 24 weeks was similar to that at baseline in the placebo group at all time points and significantly higher than baseline only at the 15-min time point in all groups receiving nateglinide.

At baseline, the fasting rate of insulin secretion was 115 (57) pmol · min⁻¹· m⁻² in the whole cohort. If maintained for 24 h, this rate would translate to an output of 54 (26) units of insulin (range 11–179 units). Fasting rates of insulin secretion did not differ among groups at baseline and showed no significant change from baseline at 24 weeks in any of the treatment groups (Table 2). At baseline, total insulin release during the 4-h postprandial period averaged 56 (26) nmol/m (18 [9] units) in the whole group, which represented a doubling of fasting insulin output (9 units over 4 h). Total insulin output did not differ among treatment groups at baseline nor did it show significant changes following nateglinide treatment (Table 2).

Thus, treatment-induced changes in postprandial glucose levels could not be explained by absolute changes in either fasting or total insulin secretion. However, when changes in insulin release were related to the concomitant changes in glucose concentrations in a model of β-cell function, a different picture emerged. Firstly, rate sensitivity, which at baseline averaged 0.32 (0.63) nmol · m⁻² · mmol⁻¹ · l in the whole group (Table 3), was significantly enhanced at 24 weeks with the lowest nateglinide dose (30 mg), with no further stimulation at higher doses (Fig. 1A). Early potentiation (1.12 [0.25] at baseline) increased with increasing nateglinide dose until it reached a value (1.35 [0.84] with 120 mg nateglinide) that was significantly higher than that achieved in the placebo group (1.20 [0.34]) (Fig. 1B). On the other hand, the β-cell dose response was steeper and shifted upwards with the 120-mg dose (Fig. 2). The dose-response parameters (Figs. 1C and D) were affected by treatment, and β-cell glucose sensitivity, which at baseline averaged 45 ± 51 pmol · min⁻¹ · m⁻² · mmol⁻¹ · l, at 24 weeks showed a trend toward increasing with increasing nateglinide dose, the change being statistically significant at the 120-mg dose. A similar trend was observed for insulin secretion at 7 mmol/l glucose, which was similar across groups at baseline (Table 3) and significantly different from placebo at the highest dose.

The changes in insulin sensitivity, as assessed by homeostasis model assessment, had a large variance and were not significant; a slight decrease (∼5%) of the OGIS index was observed in the placebo group, which was significant in comparison to the highest dose (Table 2).

At baseline, postprandial glucose levels were closely related to glucose sensitivity in a reciprocal manner (r = −0.57, P < 0.0001, after log-log transformation). To assess the independent contribution of parameters of β-cell function to glucose tolerance, the treatment-induced changes in plasma glucose concentrations in the whole dataset were regressed simultaneously against the changes in β-cell parameters. Higher baseline glucose levels (whether fasting, postprandial, or incremental) were associated with greater decrements of the corresponding glucose levels at 24 weeks. Independently of baseline values, the treatment-induced changes in rate sensitivity, glucose sensitivity, and potentiation all contributed to the decrements in both absolute and incremental postprandial glucose concentrations (Table 4). In contrast, only glucose sensitivity made a statistically significant, if small, contribution to the changes in fasting plasma glucose levels.

In in vitro and in vivo systems, nateglinide stimulates insulin release in a manner that is glucose dependent, of rapid onset, and rapidly reversible (10). The expected in vivo effect of chronic nateglinide treatment is therefore an enhancement of early insulin release. In the present study, this phenomenon was barely visible when using traditional methods of measuring insulin secretion (Table 2). Thus, the insulinogenic index, a marker of the early insulin response, was not found to be significantly increased. On the contrary, the effects of nateglinide on β-cell function were clearly apparent using the model. Rate sensitivity, which is the β-cell function parameter associated with early insulin secretion, was increased at all drug doses (Fig. 1). The positive effect of nateglinide on rate sensitivity is noteworthy, as rate sensitivity is markedly depressed in patients with type 2 diabetes (6,8). Rate sensitivity accounts for a very small fraction (1.7 [2.9] nmol/m or 3.5 ± 0.5% in the whole cohort at baseline) of the total amount of hormone released during 4 h following meal ingestion. This fraction increased by an average of 2.8 ± 0.9% (P = 0.002 vs. placebo) with active treatment (mean of all three doses) and was associated with a reduction in postprandial glucose levels of 1.5 ± 0.4 mmol/l and a decrease in postprandial glycemic excursions of 0.8 ± 0.3 mmol/l (P < 0.01 vs. placebo for both). These quantitative findings highlight the critical importance of rate sensitivity for the maintenance of tolerance to ingested carbohydrate: a prompt secretory response is key to the subsequent control of glucose levels by insulin. Also of note is that nateglinide treatment was associated with a rather large improvement in rate sensitivity (from 0.32 [0.63] to 0.76 [0.93] nmol · m⁻² · mmol⁻¹ · l⁻¹, all doses) to a value not too far from that observed in nondiabetic subjects (1.4 ± 1.3 nmol · m⁻² · mmol⁻¹ · l⁻¹ [7]). Restoring rate sensitivity during the early phase of the secretory response to a meal may be critical to rapidly inhibit endogenous glucose production, thereby limiting postprandial glucose excursions (14). It is relevant in this regard that in the present study, insulin sensitivity, as estimated by homeostasis model assessment or OGIS, was not affected by nateglinide. Therefore, the improvement in glucose tolerance with nateglinide appears to be solely determined by the improvement of β-cell function. This result, obtained from an intervention study, confirms the importance of β-cell function as a determinant of glucose tolerance, as we have previously observed in cross-sectional clinical studies (7,8).

Early potentiation, glucose sensitivity, and insulin secretion at 7 mmol/l glucose also improved following nateglinide treatment; the dose dependence of these changes was, however, different from that of rate sensitivity. In fact, although an upward trend for these two indexes was evident across nateglinide doses (Fig. 1), statistical significance was achieved only with the highest dose. The most important consequence of increasing nateglinide dose is to prolong the time period over which the drug can enhance insulin secretion (10,15). At the highest dose (120 mg), almost all patients are expected to maintain adequate drug levels to enhance insulin secretion over the entire mealtime period. At lower doses, the nateglinide effect on insulin secretion may have waned due to inadequate drug levels before glucose levels had returned to baseline.

While a direct effect of nateglinide on early potentiation, β-cell glucose sensitivity, and insulin secretion at 7 mmol/l glucose is possible, it must be noted that treatment with the highest drug dose was associated with significant chronic decrements in both fasting and postprandial glucose levels, which by themselves may have enhanced glucose sensitivity and potentiation. A study protocol comparing nateglinide with another hypoglycemic agent achieving equivalent glycemic control would answer the question whether abatement of glucose toxicity can enhance β-cell activity regardless of how glucose levels are lowered.

Finally, it is of considerable interest that all three model-derived measures of β-cell function were significant independent predictors of glucose tolerance in a prospective sense, i.e., their improvements predicted the declines in postprandial hyperglycemia observed over a 6-month period (Table 4). Although the quantitative contribution was modest, this finding does provide an important validation for the model-based description of β-cell function.

In conclusion, this study demonstrates that in mildly hyperglycemic patients with type 2 diabetes, several model-calculated parameters of dynamic insulin release can be improved by 24 weeks of treatment with nateglinide. This improvement occurs at no expense in terms of β-cell stress, namely, without any change in the total amount of hormone secreted in response to a meal challenge.

This work was supported by funding from Novartis Pharmaceutical, a European Foundation for the Study of Diabetes-Novo Nordisk Type 2 Programme Focused Research Grant, and funds from the Italian Ministry of University and Scientific Research (MURST prot. 2001065883_001).

The authors gratefully acknowledge Shamita Gupta, Patricia Rumpelt, Michele Ball, and David Holmes for clinical study coordination. Additionally, we thank the investigators and study personnel at the participating clinical centers in Argentina, Australia, Belgium, Canada, Finland, France, Germany, Italy, the Netherlands, New Zealand, Sweden, and the U.S. for their contributions to the study.