In health, insulin is secreted to precisely meet metabolic demand and maintain circulating glucose concentration within strict limits. The rates of hepatic glucose production and glucose utilization are regulated by circulating insulin levels, remaining closely matched and relatively constant from day to day. Endogenous insulin secretion is stimulated primarily by increases in blood glucose levels. Non-glucose initiators of insulin secretion include certain monosaccharides, amino acids and fatty acids [11, 22]. It has been suggested that insulin secretion may also be influenced by other factors acting directly via cell surface receptors [23]. The discovery of G-protein fatty acid receptors that are either directly (GPR40) or indirectly (e.g. GPR120) involved in insulin secretion has generated new targets for diabetes drug development (see below) [24].

Glucose is transported from the blood into the β-cell via the high-capacity glucose transporter (GLUT2) system. Inside the cells, it is metabolized by glucokinase to generate glucose-6-phosphate [25]. Glycolytic and oxidative metabolism of glucose elevates the cytosolic adenosine triphosphate/adenosine diphosphate (ATP/ADP) ratio. This closes ATP-sensitive potassium channels in the cell membrane. The resulting depolarization triggers opening of voltage-gated calcium channels. Increased intracellular calcium concentration allows the fusion of insulin-containing granules with plasma membrane and the subsequent release of stored insulin through interactions with Ca²⁺-sensitive proteins (Fig. 2.1).

Even in the unstimulated (basal) state, insulin secretion is continually tuned by numerous stimulatory and inhibitory signals [26]. Insulin secretion is influenced by central neural activity [27] and paracrine factors [28, 29]. Autocrine regulation of insulin secretion has been proposed [30, 31], although the physiological relevance of such a mechanism in humans has been questioned [32].

Glucose-stimulated insulin secretion to a rapid and sustained increment in plasma glucose is characteristically biphasic (Fig. 2.2) [33]. A transient first phase of insulin secretion with an early peak is followed by a gradually developing and more prolonged second phase. It has been suggested that the first phase may represent release of insulin from rapidly mobilized storage granules, triggered by the influx of Ca²⁺; the second phase of insulin secretion may reflect ATP-dependent recruitment and release of granules from a reserve pool [5, 34]. The relevance of first-phase insulin secretion to glucose homeostasis in human physiology remains uncertain [35].

At matched plasma glucose concentrations, orally ingested glucose elicits a substantially greater insulin secretory response than intravenous glucose; this is the so-called incretin effect [36]. The main incretin hormones, GLP-1 and glucose-dependent insulinotropic polypeptide (GIP), are released from intestinal endocrine cells (L cells and K cells, respectively) within minutes of nutrient ingestion. The potentiation of glucose-stimulated insulin secretion by the incretin hormones may account for approximately 50–70 % of postprandial insulin secretion [37, 38]. GLP-1 and GIP circulating in the blood are rapidly inactivated via enzymatic cleavage by dipeptidyl peptidase 4 (DPP-4) [39, 40]. GLP-1 and GIP act via specific G-protein-coupled cellular receptors to raise intracellular levels of cyclic adenosine monophosphate (AMP) and inhibit ATP-sensitive K⁺ channels, actions which induce β-cell exocytosis of insulin (see Fig. 2.1) [41]. Incretin-mediated augmentation of insulin secretion and inhibition of glucagon secretion are dependent on ambient glucose concentrations [42]. In addition, GLP-1 reduces gastric emptying, increases satiety and reduces food intake [43] as well as stimulating the transcription of glucokinase and the GLUT 2 transporter gene [44].

Defects in incretin axis physiology are implicated in the pathogenesis of type 2 diabetes. Delineation of these defects has led to the development of important new classes of glucose-lowering drugs, i.e. the DPP-4 inhibitors and GLP-1 analogues [40]. Reported reductions in GLP-1 responses in patients with type 2 diabetes provided the theoretical basis for these incretin-based glucose-lowering drugs [45]. The impact of hyperglycaemia and incretin effect on insulin secretion and β-cell function continues to be unravelled [46, 47]. In a placebo-controlled study of subjects (n = 22) with impaired fasting glucose, sitagliptin increased intact circulating GLP-1 concentrations but had no effect on postprandial plasma glucose, insulin or C-peptide concentrations [48]. While the secretion of GIP is near normal in patients with type 2 diabetes, the effect of this incretin on insulin secretion, particularly the late phase, is impaired [38].

Insulin and C-peptide are co-secreted into blood that is directed to the portal vein, and insulin directly regulates hepatic glucose metabolism [16]. Hepatic first-pass clearance of insulin removes >50 % of insulin ensuring that the portal:systemic insulin gradient is 2:1 or higher. Insulin clearance, of which hepatic extraction is the main determinant, may be altered in common metabolic diseases [49] and possibly by certain glucose-lowering drugs [50]. Since it is not cleared by the liver, the serum concentration of C-peptide in the peripheral circulation provides a more reliable indicator of endogenous insulin secretion. The most robust assessment of dynamic β-cell insulin secretion would require blood samples drawn directly from the hepatic portal vein. Such a highly invasive approach is not feasible in clinical research, although the technique has been applied in nonhuman primate studies [51]. Prehepatic insulin secretion may be estimated by mathematical deconvolution of serum C-peptide in peripheral blood (see below) [52]. The half-life of serum C-peptide is longer than that of insulin (~20–30 vs. ~3–5 min). During dynamic tests of insulin secretion, the longer half-life in the circulation of serum C-peptide requires that blood sampling time points for measurement of C-peptide be adjusted relative to those for insulin. Impaired renal function reduces C-peptide clearance [53].

Selective immune destruction of β-cells is the principal pathogenic defect of type 1 diabetes [66]. The resulting insulin deficiency necessitates lifelong replacement therapy with subcutaneous administration of insulin or insulin analogues (or a successful pancreas or islet transplant). However, recent data suggest that even patients with type 1 diabetes of long duration may have low levels of residual insulin secretion [67, 68].

In clinical trials of patients with type 1 diabetes, β-cell function is commonly assessed by the serum C-peptide response to intravenous glucagon [69] or a mixed meal [70]. Glucagon is injected as an intravenous bolus with timed measurements of serum C-peptide [69]. The glucagon stimulation test may also be of value in clinical practice when therapeutic decisions may be are informed by assessment of endogenous insulin secretion [71].

Islet β-cell insufficiency is central to the development and progression of type 2 diabetes. A decline in β-cell function antedates and predicts the onset of clinical diabetes [61]. Genetic and environmental factors are strongly implicated [55]. Characterization of defective insulin secretion in the precursor states of impaired fasting glucose (IFG) and impaired glucose tolerance (IGT) has provided valuable insights into the pathogenesis of type 2 diabetes [72]. In the United Kingdom Prospective Diabetes Study (UKPDS), β-cell function, determined using a modelling method (HOMA-B, see below), was reduced by approximately 50 % in middle-aged subjects at the time of diagnosis [73]. This is in line with estimates based on autopsies place the reduction of β-cell mass at approximately 40–65 % at the time of clinical presentation of type 2 diabetes [74]. A reduction in functional β-cell mass is evident during the long preclinical phase of type 2 diabetes. In parallel, the regulation of blood glucose both postprandially and in the fasting state deteriorates through stages of glucose intolerance before rising to levels diagnostic of diabetes. Insulin secretion rates rise through the range between normal glucose tolerance to IGT and decline thereafter [75]. Longitudinal studies have demonstrated progressive and variable, but not inevitable, loss of β-cell function in patients with type 2 diabetes [76, 77]. As the number of β-cells per islet falls, plaques of amyloid fibrils, which may contribute to β-cell loss, are deposited [78, 79]. Hyperglycaemia (glucotoxicity), both acute [80] and chronic [81], and prolonged elevations of plasma concentration of nonesterified (‘free’) fatty acid concentrations (lipotoxicity) [82] are detrimental to β-cell function and viability [10, 83]. Oxidative and endoplasmic reticulum stress together with inflammation have been identified as pathways to β-cell loss that are potentially amenable to targeted therapeutic interventions [84]. As hyperglycaemia develops, insulin pulses are attenuated [85] and circulating levels of proinsulin and partially processed proinsulin products become are elevated, albeit remaining at low concentrations relative to insulin [86]. Loss of acute glucose-stimulated insulin secretion is accompanied by alterations in β-cell phenotype and in gene and protein expression.

Uncommon monogenic forms of diabetes in which β-cell defects have been well delineated, e.g. forms of maturity-onset diabetes of the young (MODY) [87, 88], contrast with the complexity of type 2 diabetes in which the relative contributions of genetic, epigenetic and nongenetic factors to β-cell failure remain uncertain [61, 89]. To date, the utility of molecular genetics in identifying personalized therapeutic approaches for diabetes has been confined to relatively small subgroups of patients. Diabetes due to hepatic nuclear factor-1α mutations is especially sensitive to sulphonylureas [90]. Neonatal diabetes due to rare activating mutations in genes encoding the ATP-sensitive potassium channel subunits Kir6.2 or SUR1 can be successfully treated with high-dose sulphonylureas [90].

The causes of type 1 are complex and multifactorial, involving genetic and environmental factors [91]. The natural history of type 1 diabetes involves a prolonged and variable latent preclinical period that culminates in the destruction of pancreatic β-cells [92]. When autoimmune-mediated β-cell loss reaches a critical point, hyperglycaemia develops. A temporary partial recovery of β-cell function may occur after initiation of insulin therapy – the so-called honeymoon period – before waning again thereafter. A diverse range of strategies for averting type 1 diabetes (primary prevention) or preserving β-cell function in subjects with recently diagnosed type 1 diabetes (secondary prevention) have been explored in clinical trials [93]. Examples of interventions directed at arresting the immune process include nicotinamide, insulin injections, oral insulin, nasal insulin, glutamic acid decarboxylase, cyclosporine, teplizumab and abatacept [93]. These challenging intervention studies have been aided by the development of biomarkers of autoimmunity [94] and serial tests of β-cell function [95]. The latter include the mixed meal tolerance test (see below) and glucagon-stimulated serum C-peptide response; these methods should not be assumed to provide equivalent results [96]. Recent research has shown that endogenous glucagon dynamics are also altered early in the course of type 1 diabetes with elevated levels in response to a mixed meal challenge [97]. The plurality of aforementioned therapeutic targets attests to the complex pathophysiology of β-cell attrition in type 1 diabetes. Novel non-immunomodulatory strategies are being explored. For example, preclinical evidence suggests that verapamil, a calcium channel blocker used for the treatment of hypertension and cardiac arrhythmias, may enhance β-cell survival and function [98]. In a recent study of patients with newly diagnosed type 1 diabetes the lipid-modifying agent atorvastatin appeared to slow the decline of β-cell function (stimulated C-peptide secretion), in a subgroup defined by higher levels of inflammation-associated immune mediators [99].

In obese subjects with glucose intolerance behavioural interventions that indirectly improve insulin secretion reduce the risk of progression to diabetes [100]. Many drugs have been shown to increase insulin secretion in vitro, but few have therapeutic potential [101]. Thiazolidinediones, which do not directly stimulate insulin secretion, have also been shown to improve β-cell function and reduce the risk of new-onset diabetes in insulin-resistant women with a history of gestational diabetes [102]. The latter observation illustrates the pathophysiological relevance of the feedback loop between insulin secretion and insulin action [103]. Lifestyle modification is the cornerstone of controlling hyperglycaemia in established type 2 diabetes [104]. Bariatric surgery procedures [105] or voluntary dietary calorie restriction of a sufficient degree can rapidly lower blood glucose concentrations and improve β-cell function [106]. A doubling of β-cell function in obese patients with type 2 diabetes 1 year after gastric bypass surgery has been reported [107]. When standard behavioural changes prove insufficient, pharmacotherapy is required to control hyperglycaemia. Sulphonylureas have the mainstay of pharmacological strategies to enhance insulin secretion for decades [108]. Agents in this class act primarily by blocking ATP-sensitive potassium channels in the β-cells thereby stimulating insulin release [109]. Dipeptidyl peptidase (DPP)-4 inhibitors, which potentiate insulin secretion in a glucose-dependent fashion by reducing the breakdown of GLP-1, have superior safety and tolerability profiles to sulphonylureas [110]. Unlike sulphonylureas, which continue to stimulate insulin secretion even during hypoglycaemia [111], DPP-4 inhibitors and GLP-1 analogues promote insulin exocytosis only when blood glucose levels are elevated. This property reduces the risk of hypoglycaemia when these drugs are used as monotherapy or in combination with other agents that have a low propensity to cause hypoglycaemia [112, 113]. Preclinical data suggesting effects of DPP-4 inhibitors on functional β-cell mass and pancreatic insulin content in rodent models raised hopes that the natural history of type 2 diabetes might be modified [114]. The UKPDS demonstrated an inexorable loss of β-cell function with sulphonylureas and insulin [73]. As yet, however, there is no firm evidence from clinical studies that GLP-1 agonists [115] or DPP-4 inhibitors [114] have durable effects on β-cell function. Progressive loss of β-cell function is regarded as the main reason that many patients with type 2 diabetes ultimately require insulin replacement therapy [73]. The hypothesis that β-cell function may be preserved by using specific combinations of glucose-lowering drugs is being tested in clinical trials [116]. New drugs designed to improve β-cell function that have entered clinical trials in recent years include:

In parallel, potential cardioprotective actions of GLP-1, highly attractive in the context of the enhanced risk of cardiovascular disease in type 2 diabetes, have also been explored [126]. A novel hypothesis that extends to the broader cardiometabolic risk profile of patients with type 2 diabetes is that strategies to raise high-density lipoprotein (HDL) levels may improve β-cell function [127]. The cholesteryl ester transfer protein inhibitor torcetrapib improved glycaemic control in atorvastatin-treated patients with type 2 diabetes. However, it remains to be determined whether this effect was a direct consequence of raising HDL levels [128]. Further elucidation of the molecular regulation of β-cell mass and function offers the prospect of additional therapeutic agents to preserve insulin secretion [129].

In the design of studies of novel therapies for diabetes, the potential for confounding by drugs that have effects on β-cell function should be carefully considered (Tables 2.1 and 2.2). In studies of new pharmacological entities for type 2 diabetes, individuals who are treatment naïve or on stable metformin monotherapy are preferred (see Chap.​ 9). Washing off sulphonylureas and DPP-4 inhibitors is a routine practice in early-phase studies of diabetes drugs although careful monitoring of glycaemic control is required according to regulatory guidelines [130]. When thiazolidinediones [131] or GLP-1 analogues [132] are temporarily withdrawn, a sufficient length of time is required to ensure that the any metabolic effects of these drugs have fully dissipated [130].

Aside from glucose-lowering medications, many patients with diabetes are treated with antihypertensive and lipid-modifying drugs that may have class-specific effects to either increase or decrease endogenous insulin secretion [133, 134]. Recent reports of adverse effects of statins on glucose metabolism are pertinent in this context [135]. Certain statins may affect insulin secretion through direct and/or indirect effects on β-cell calcium channels [136].

The feedback loop between the β-cells and insulin-responsive tissues requires that circulating insulin concentrations be interpreted in the context of the prevailing blood glucose level. Changes in both insulin sensitivity and β-cell function are evident as glucose tolerance deteriorates [149]. When fasting hyperglycaemia is present, the predicted level of hyperinsulinaemia would be higher if β-cells were able to fully overcome any defect in insulin action through compensatory increased insulin secretion [150]. Demonstration of an effect on residual β-cell function is an important consideration in studies of new drug therapies for type 2 diabetes. This is most readily accomplished by measurement of serum or plasma C-peptide after an overnight fast. Several classes of diabetes drugs either stimulate insulin secretion (e.g. sulphonylureas, DPP-4 inhibitors) or require the presence of sufficient insulin to exert their glucose-lowering effects (e.g. metformin, thiazolidinediones) [108]. Exceptions include α-glucosidase inhibitors [151] and sodium glucose cotransporter 2 (SGLT2) inhibitors [152].

The use of insulin-specific immunoradiometric assays eliminates cross-reactivity with proinsulin and partially processed proinsulin molecules [153]. Loss of the normal pulsatile oscillations in insulin secretion [85] and increased ratios of circulating proinsulin to insulin [153, 154] are regarded as markers of compromised β-cell function in prediabetic states of glucose intolerance. In patients with type 2 diabetes, the ratio of proinsulin to insulin is inversely related to β-cell function [155]. In studies of an insulin secretagogue and an insulin sensitizer in patients with type 2 diabetes, a differential effect of the two classes of glucose-lowering drugs was observed on plasma proinsulin and the proinsulin to immunoreactive insulin ratio [155]. Glibenclamide (glyburide) increased, whereas rosiglitazone decreased, the concentration of proinsulin and the proinsulin-to-insulin ratio [155]. In a 12-month head-to-head study of these two drugs, rosiglitazone reduced the concentrations of insulin, proinsulin and split proinsulin compared to glibenclamide [156]. As expected, rosiglitazone improved insulin sensitivity, whereas glibenclamide therapy was associated with worsening insulin resistance [156]. These results were replicated in ADOPT (A Diabetes Outcome Progression Trial) [157]. Over the 4-year course of ADOPT, rosiglitazone had more favourable effects on β-cell function than glibenclamide [158]. The improvements in β-cell function with rosiglitazone, which were accompanied by a decrease in insulin resistance, were associated with a more durable effect on glycaemic control [157, 158]. Whether the beneficial effects of rosiglitazone on β-cell function were indirect via improved insulin sensitivity or resulted from a direct β-cell action [159] could not be determined from this study (see below). Metformin, which was also studied in ADOPT, had effects on β-cell function and glycaemic control that were intermediate to those observed with rosiglitazone and glibenclamide [157, 158].

Homeostasis model assessment is a method for assessing insulin secretory capacity from fasting plasma insulin and glucose concentrations. HOMA-B has been widely applied in clinical metabolic research, including diabetes drug development.
An early study of diabetic and nondiabetic subjects reported a coefficient of variation (CV) for HOMA-B measured on two occasions of approximately 30–40 % [160]. Subsequent studies using specific insulin assays in larger samples have yielded CVs closer to 10 % [138]. The use of triplicate samples may improve the CV with smaller sample sizes [138]. In comparative studies, HOMA-B has been shown to correlate with measures of insulin secretion obtained using the hyperglycaemic clamp (see below) in diabetic (r = 0.59) and nondiabetic subjects (r = 0.71) [160]. Since HOMA values are rarely normally distributed, the data should be logarithmically transformed and presented as geometric means with appropriate measures of dispersion.

The HOMA model apportions the basal state of insulin and glucose in terms of insulin resistance and β-cell function [138]. Insulin sensitivity and insulin secretion are mutually related as a hyperbolic function in which insulin resistance is compensated by increased insulin secretion [54, 103]. Reporting HOMA-B data in isolation, i.e. without considering the prevailing level of insulin sensitivity for the individual, could lead to erroneous conclusions about β-cell function [103, 138]. The validity of assumptions inherent in HOMA-B has been challenged [161]. It has been argued that measures of β-cell function and insulin sensitivity should be as independent as possible; clearly, HOMA does not satisfy this requirement [103]. HOMA-B is a measure of β-cell activity rather than health; accordingly, the temporary improvement observed in response to a classic secretagogue, i.e. a sulphonylurea, should be regarded as reflecting the mechanism of action of the drug rather than any intrinsic change in functional capacity [73].

HOMA-B provides an index of insulin secretory capacity in the non-stimulated state at serum insulin levels primarily of relevance to hepatic glucose regulation [162]. This focus may detract from the ability of HOMA-B to quantify defective β-cell function relative to dynamic tests. In the Insulin Resistance Atherosclerosis Study (IRAS), HOMA-B underestimated the magnitude of the β-cell defect in subjects with IGT and newly diagnosed diabetes compared with the acute insulin response assessed using an intravenous glucose challenge (see below) [163]. In a study in which insulin secretion was compared with pancreatic histology, the glucose-stimulated C-peptide-to-glucose ratio appeared to predict β-cell area better than HOMA [164]. With these caveats in mind, HOMA-B offers a simple and useful approach for assessing β-cell function that can be of value in evaluating the actions of new diabetes drugs. For example, the insulin sensitizer pioglitazone lowers blood glucose in patients with type 2 diabetes primarily by improving insulin action [131]. Thiazolidinediones also improve β-cell function through indirect mechanisms, an effect demonstrable as an increase in HOMA-B [165]. For DDP-4 inhibitors, which primarily enhance postprandial insulin secretion, an oral glucose challenge would be more logical than an assessment of fasting β-cell function. However, in a recent meta-analysis of placebo-controlled clinical trials of sitagliptin, improved β-cell function was evident as an increased HOMA-B index [166]. Note that HOMA-B cannot be used in patients treated with exogenous insulin.

The 75 g oral glucose tolerance test (OGTT) is the reference method for diagnosing diabetes and other categories of glucose dysregulation [167]. Thus, regression from one diagnostic category to another, e.g. IGT to normal glucose tolerance, in response to therapy can be determined. For example, in the DREAM (Diabetes Reduction Assessment with Ramipril and Rosiglitazone Medication) study in adults with IFG and/or IGT, rosiglitazone significantly reduced the incidence of type 2 diabetes and increased the likelihood of regression to normoglycaemia compared to placebo [168].

The OGTT can also provide an integrated assessment to the β-cell response to an intestinal glucose stimulus that includes activation of the incretin axis. As discussed above, the incretin effect on the islets includes enhancement of glucose-stimulated insulin secretion in concert with reduced glucagon release [169]. When assessing the response to therapeutic agents whose primary mode of action is mediated via the incretin axis, e.g. DDP-4 inhibitors, the response of the major incretin hormones (GLP-1, GIP) can be quantified and related to the stimulation of insulin secretion and suppression of glucagon secretion. For example, in a placebo-controlled crossover study performed in patients with type 2 diabetes, a single oral dose of sitagliptin (25 mg or 200 mg) dose-dependently inhibited plasma DPP-4 activity over 24 h [170]. Sitagliptin increased active levels of GLP-1 and GIP, increased serum insulin and C-peptide levels, decreased plasma glucagon levels and reduced the rise in blood glucose during a 75 g OGTT [170].

For paired comparisons, the area under the curve (AUC) for serum insulin (or C-peptide) offers a simple exploratory assessment of the change in endogenous insulin secretion in response to a therapeutic intervention that incorporates early and later phases of insulin release. Consideration should be given to the methods used for data analysis in this scenario [171]. Since the OGTT is not standardized in terms of attained blood glucose concentrations, methods for normalizing insulin secretion to the glucose stimulus have been developed. A number of empirical indexes of modelling approaches have been proposed which relate insulin responses to increments in blood glucose. For example, the insulinogenic index, which reportedly correlates with first-phase insulin release assessed using an intravenous glucose bolus, is calculated as the ratio of the increment of insulin to glucose above baseline 30 min (Δinsulin/Δglucose) after an oral glucose challenge [172]. In ADOPT [157] (see above), the insulinogenic index was used to study the β-cell responses to rosiglitazone, glibenclamide and metformin. A higher acute change in the insulinogenic index was observed over the first 6 months with glibenclamide than with rosiglitazone or metformin [158]. Thereafter, glibenclamide was associated with a significantly faster rate of decline in insulinogenic index versus rosiglitazone (11.1 vs. 6.0 % per year), the decline with metformin being intermediate. Despite the fact that glibenclamide is a classic insulin secretagogue, the mean insulinogenic index with glibenclamide was lower than those in the other treatment groups beyond 2 years [158]. Patients who failed monotherapy (with any agent) had a lower median insulinogenic index at baseline compared to those who completed the 4-year trial [158]. High within-subject variability for the insulinogenic index has been reported [173] which should be considered in the design of studies with smaller sample sizes.