Impaired beta cell glucose sensitivity and whole-body insulin sensitivity as predictors of hyperglycaemia in non-diabetic subjects
- First Online:
- Cite this article as:
- Walker, M., Mari, A., Jayapaul, M.K. et al. Diabetologia (2005) 48: 2470. doi:10.1007/s00125-005-0004-7
- 401 Downloads
The aim of this prospective study was to investigate predictors of deteriorating glucose tolerance in subjects of British extraction.
A total of 156 non-diabetic subjects (86 with a family history of type 2 diabetes) underwent a 75-g OGTT and anthropometric assessment at baseline and 5 years later. Pancreatic beta cell function and whole-body insulin sensitivity were studied by model assessment. Subjects were classified as progressors if glucose tolerance moved one or more steps from normal, impaired fasting glucose, impaired glucose tolerance and diabetes over the follow-up period.
At baseline, the progressors (n=22) had increased adiposity and a higher proportion of familial diabetes and abnormal glucose tolerance than non-progressors. Baseline pancreatic beta cell sensitivity to changes in glucose (p<0.02) and whole-body insulin sensitivity (p<0.0001) were decreased in the progressors. Logistic regression revealed that baseline and follow-up changes in beta cell glucose sensitivity and insulin sensitivity, rather than the classical clinical predictors (adiposity, familial diabetes and glucose levels), were the key independent predictors of progression (explaining over 50% of the progression).
Impaired pancreatic beta cell glucose sensing and whole-body insulin sensitivity predict progression to hyperglycaemia. Strikingly, these pathophysiological changes override the importance of the clinical risk factors and highlight potential metabolic targets for prevention strategies.
KeywordsBeta cell glucose sensitivity Disposition index Impaired glucose tolerance Impaired fasting glucose Insulin secretion Insulin sensitivity Oral glucose tolerance test
World Health Organization
We are in the midst of a world-wide explosion in the prevalence of type 2 diabetes  that is the result of life-style changes on a background of genetic susceptibility. A major concern is that this will translate into a future burden of diabetes-related complications and premature mortality . Understanding the pathogenesis of type 2 diabetes will facilitate the development of future effective strategies for prevention.
A number of longitudinal studies have been conducted among different ethnic groups to identify predisposing factors. Insulin insensitivity, frequently linked to increased total and central adiposity, is a key predisposing factor [2, 3]. However, impaired pancreatic beta cell function is an essential element in the development of abnormal glucose tolerance. Indeed, altered beta cell function is a known predisposing factor for type 2 diabetes [4, 5, 6, 7, 8], while other studies have found that insulin insensitivity and abnormal beta cell function are independent predictors [9, 10, 11]. A major problem until now has been the lack of detailed information about the specific abnormalities of beta cell function that predispose to diabetes. This is in part because the methods required to investigate beta cell function—such as the measurement of insulin pulsatility, beta cell glucose sensitivity (sensitivity to changes in blood glucose levels) and the response to different secretagogues—invariably involve complicated and time-consuming protocols. As a consequence, they are difficult to apply across the large number of individuals generally needed for prospective studies of predictors of diabetes.
What is needed, therefore, are alternative methods for assessing beta cell function that on the one hand are experimentally simple, but on the other also look at different aspects of beta cell function, so that the best predictors of progression can be identified. This in turn will inform the development of targeted therapies to prevent progression. We have recently developed a mathematical model that uses data from a standard oral glucose tolerance test to measure insulin secretion rates, beta cell glucose sensitivity and beta cell potentiation (effects of glucose itself and non-glucose secretagogues to magnify the response to glucose), together with an estimate of whole-body insulin sensitivity [12, 13, 14]. In this study we used this model to identify predictors of deterioration in glucose tolerance in a longitudinal study of non-diabetic subjects of British extraction with or without a family history of type 2 diabetes.
Subjects and methods
Non-diabetic subjects of British extraction were recruited and underwent investigation at baseline (1991–1993) and 5 years later (1996–1998). Of the original 180 subjects, 156 had a full data set available for analysis. Of these 156 subjects, 92 had a first-degree relative with type 2 diabetes and the remainder had no family history of diabetes. All subjects were residents of Newcastle and North Tyneside. Informed written consent was obtained from all subjects and the study was approved by the Newcastle and North Tyneside Health Authority and University of Newcastle upon Tyne Joint Ethics Committee.
All subjects underwent the same anthropometric and metabolic investigations at baseline and approximately 5 years later. Subjects fasted for 10 h overnight, having avoided alcohol and vigorous exercise for the preceding 24 h. The following morning, an indwelling cannula was placed in an antecubital vein. Fasting samples (three samples separated by 5-min intervals) were taken, and then the subjects were given a standard 75 g oral glucose load (388 ml degassed Lucozade; Glaxo Smithkline, Brentford, UK). Further blood samples were taken 0, 10, 20, 30, 45, 60, 90 and 120 min after the oral load. Height, weight and waist and hip circumferences were measured according to the protocol of the World Health Organization (WHO) MONICA Project . Blood pressure was measured according to the guidelines of the British Hypertension Society .
Plasma glucose concentrations were measured using a glucose oxidase method (interassay CV 3.5%; YSI 2300 Stat Plus; Yellow Springs Instruments, Farnborough, UK). Sera for hormone determinations were separated at 4°C and stored at −40°C. Subject samples were analysed in batches to minimise interassay variation. Plasma insulin and C-peptide concentrations were determined by specific enzyme immunoassays (Dako Diagnostics, Ely, UK). For the insulin assay, the interassay CV was 4.8% at 34 pmol/l with 0.3 and 0% cross-reactivity with proinsulin and C-peptide, respectively. For the C-peptide assay, the inter-assay CV was 3.5% at 160 pmol/l, with 63 and 0% cross-reactivity with proinsulin and insulin, respectively. Assays for measurement of glucose, C-peptide and insulin were identical at baseline and follow-up, and throughout this period the assays were monitored as part of the UK National External Quality Assurance Scheme.
Glucose tolerance status was classified according to 2000 WHO criteria . Thus, at the baseline examination subjects were classified as having NGT (fasting glucose <6.1 mmol/l, 2-h glucose <7.8 mmol/l), IGT (fasting glucose <6.1 mmol/l, 2-h glucose 7.8–11.1 mmol/l), IFG (fasting glucose 6.1–7.0 mmol/l, 2-h glucose <7.8 mmol/l), or combined IFG and IGT (IFG*IGT). Those subjects classified as having IGT had a repeat OGTT (fasting and 2-h glucose concentrations measured only) within 2–4 weeks to identify those with persistent IGT. Based on the OGTT performed at follow-up, subjects were classified as progressors if their glucose tolerance deteriorated (from a lower step to any of the higher steps in the ascending sequence NGT → IFG → IGT → IFG*IGT → diabetes) or non-progressors if their glucose tolerance status did not change or improved.
Insulin sensitivity, as an estimate from the OGTT of the glucose clearance during a euglycaemic clamp (expressed in ml per min per square metre of body surface area), was calculated from the plasma glucose and insulin values during the OGTT according to the method of Mari and colleagues , which has been validated against the euglycaemic insulin clamp technique. Insulin secretion rates were calculated from plasma C-peptide concentrations by deconvolution, as previously described . Parameters of beta cell function were derived from mathematical analysis of plasma glucose and C-peptide concentrations during the OGTT, using a previously developed model [13, 14]. According to this method, glucose-stimulated insulin secretion [S(t)] is the sum of two components: S(t)=P(t)f(G)+Sd(t). The first component, P(t)f(G), represents the dependence of insulin secretion on the absolute glucose concentration (G) at any time point. It is expressed as the product of a dose–response function, f(G), relating insulin secretion to glucose concentration, and a potentiation factor, P(t), which modulates the dose–response relationship. The potentiation factor represents a time-dependent relative augmentation of insulin secretion [14, 19] due to several factors (glucose-mediated potentiation, non-glucose secretagogues, gastrointestinal hormones and neurotransmitters). Thus, P(t) is a positive function of time and averages 1 during the experiment.
A characteristic parameter of the dose–response curve is its mean slope in the observed glucose concentration range, denoted beta cell glucose sensitivity. Glucose sensitivity represents the ability of the beta cell to respond to an increment in glucose concentration. It has been shown to have good reproducibility (24%) . A characteristic parameter of the potentiation factor is the ratio between its value at the end of the OGTT (100–120 min) and its initial value (0–20 min). It has been observed that this potentiation ratio is greater than 1 in normal subjects (representing a relative increase in insulin secretion in relation to the glucose levels over the course of the OGTT), and is progressively depressed as glucose tolerance worsens .
The second insulin secretion component, Sd(t), accounts for the observation that rapid changes in glucose concentration enhance insulin release. Sd(t) is proportional to the rate of change of plasma glucose concentration during the OGTT [Sd(t) = kd dG(t)/dt for dG(t)/dt > 0, otherwise Sd(t) = 0]. A characteristic parameter of this secretion component is the constant kd, termed ‘rate sensitivity’. This parameter has limited reproducibility (52%) .
The model parameters were estimated from glucose and C-peptide concentration by regularised least-squares analysis, as previously described [13, 14]. Regularisation involves the choice of smoothing factors that were selected to obtain glucose and C-peptide model residuals with standard deviations close to the expected measurement error (~2% for glucose and ~4% for C-peptide).
Mean plasma glucose and insulin concentrations during the OGTT were calculated by the trapezoidal rule. The insulinogenic index was calculated as the ratio of the insulin concentration increment to the glucose concentration increment at 30 min into the OGTT (ΔI30/ΔG30). The ability of the beta cell to compensate for the extant degree of insulin resistance was estimated by calculating the product of insulin sensitivity and the insulinogenic index, analogous to the disposition index (DI, in units of ml min−1 m−2 pmol mmol−1) proposed by Kahn and colleagues .
Normally distributed data are given as the mean ± SD. Non-normally distributed data are presented as the median and interquartile range (difference between 75th and 25th percentile numbers) in the tables and text, and plotted as median and SEM in the figures. Categorical variables were compared by the chi square test, continuous variables by the Mann–Whitney U test. Changes over time were analysed by analysis of covariance, with the change in parameter (value at 5 years minus baseline value) as the dependent variable, group as the independent variable, and the baseline value as a covariate. For this analysis, non-normally distributed parameters were transformed into their natural logarithms. Logistic regression was used to test for the association between predictor variables and incident abnormalities of glucose tolerance. Results are given as the odds ratio (OR) and 95% CI. A p value of <0.05 was considered statistically significant.
Baseline clinical characteristics according to progression to hyperglycaemia
Sex, M/F (n)
BMI (kg m−2)
WHR (cm cm−1)
Fasting glucose (mmol/l)
2-h glucose (mmol/l)
Mean glucose (mmol/l)
Fasting insulin (pmol/l)
Mean insulin (pmol/l)
Baseline metabolic parameters according to progression to hyperglycaemia
Fasting ISR (pmol min−1 m−2)
Total IS (nmol m−2)
Beta cell glucose sensitivity (pmol min−1 m−2 mmol/l−1)
Rate sensitivity (nmol m−2 mmol/l−1)
Insulin sensitivity (ml min−1 m−2)
DI (ml min−1 m−2 pmol mmol−1)
Changes in glucose tolerance and metabolic measures with follow-up
Changes in metabolic data between baseline and 5 years according to progression to hyperglycaemia
Beta-cell glucose sensitivity
Predictors of progression
The present study is based on a cohort enriched with individuals at enhanced risk of incident hyperglycaemia because of a positive family history of diabetes. Numerous studies have found that individuals with a positive family history are at increased risk of developing diabetes [22, 23, 24], and show decreased insulin secretion after adjustment for whole-body insulin sensitivity in cross-sectional studies [25, 26]. In our cohort of non-diabetic subjects, there were fewer subjects with NGT in the group with familial diabetes than in the group without such a history. Over 5 years, glucose tolerance deteriorated according to the defined criteria in a total of 22 subjects, i.e. at a rate of 1.4%/year for individuals with no family history of diabetes and 4.0%/year for subjects who had a first-degree relative with diabetes (p=0.02). This underlines the increased risk of progression in those subjects with a positive family history of type 2 diabetes.
The clinical phenotype of the progressors was characterised by central obesity, slightly elevated fasting and postchallenge glucose levels, and modest whole-body insulin insensitivity. While plasma insulin concentrations, fasting or after glucose, did not clearly discriminate between progressors and non-progressors, insulin secretory rates were higher in the progressors. Nonetheless, dynamic indices of beta cell function, such as glucose sensitivity and beta cell compensation for insulin insensitivity , were significantly and markedly impaired (by ~25%) in the progressors. Thus, insulin hypersecretion masked the fact that the intrinsic ability of the beta cell to acutely respond to glucose was already compromised in the progressors at baseline.
These findings suggest that whole-body insulin insensitivity and beta cell dysfunction not only mark the high-risk condition of progressors but precede the actual deterioration of glucose tolerance in ‘confirmed’ individuals. Analysis of the prospective data supports this interpretation. In a multivariate model of prediction, baseline whole-body insulin insensitivity, beta cell glucose insensitivity and their changes over time were independent predictors of incident hyperglycaemia, together explaining a large proportion of the observed progression. A striking and novel finding of this study is that these pathophysiological changes totally replaced the classical clinical risk factors (such as obesity, familial diabetes and plasma glucose concentrations) as predictors of progression towards type 2 diabetes. This differs somewhat from the findings of a similar prospective study  in which type 2 diabetes was predicted by a combination of metabolic and classical risk factors (obesity and plasma glucose levels). However, this may be because the metabolic measures were comparatively restricted, and did not include detailed assessment of beta-cell function, such as measures of insulin secretion and glucose sensitivity.
Our results extend previous findings [9, 11] in several respects. First, whole-body insulin insensitivity, even of a mild degree and at a relatively young age, is a consistent antecedent of hyperglycaemia of any severity (ranging from IFG to overt diabetes). Second, when beta cell glucose insensitivity is measured (by modelling in our case), it proves to be a powerful determinant of glucose tolerance and an accurate predictor of its changes over time regardless of the actual amounts of insulin secreted. The disposition index assesses the adequacy of the acute insulin response to the prevailing insulin sensitivity . In our cohort, the disposition index was a weak predictor of incident hyperglycaemia in univariate association (OR=1.32; CI=0.97–1.79, p=0.08) and was not an independent predictor in the multivariate model including beta-cell glucose sensitivity and insulin sensitivity. Finally, no previous study has assessed the impact of changes in different aspects of beta cell function and insulin sensitivity over time on glucose intolerance. This was made possible in our study by the use of the OGTT, i.e. a relatively simple clinical test, as the data source for modelling analysis of both beta-cell function and insulin sensitivity. It is important to emphasise that, while whole-body insulin insensitivity and beta cell glucose insensitivity were quantitatively small in our cohort, they interact with one another strongly. Thus, insulin insensitivity and its changes over time explained 40% of the progression, and beta cell glucose insensitivity and its spontaneous changes over time explained 22% of the progression, but the two together accounted for over half of the observed variability.
We accept that this study has some limitations. The cohort was comparatively small, although it is the largest prospective longitudinal study of its kind to date of a cohort of British extraction. The classification of conversion was based upon progression to a composite based upon IFG and IGT and frank type 2 diabetes. However, evidence from the Pima Indians indicates that the pathophysiology underlying the deterioration in glucose tolerance is the same whether from NGT to IGT or from IGT to type 2 diabetes . Finally, the metabolic measures are generated from the modelling of data derived from a standard OGTT. However, the overall finding that measures of beta cell function and insulin sensitivity are the key determinants of deterioration of glucose tolerance is in keeping with the findings from similar longitudinal studies that have used other empirical methods to measure baseline metabolic functions. Importantly, our approach is suitable for large-scale longitudinal studies.
In conclusion, we have shown that changes in both pancreatic beta cell function and whole-body insulin sensitivity are associated with deterioration in glucose tolerance in subjects of British extraction. Preliminary evidence from this study suggests that changes in glucose sensing appear to be a key change in beta cell function. Strikingly, it is clear that the pathophysiology overrides the effects of the classical clinical risk factors (obesity, familial history and plasma glucose levels) in predicting deterioration in glucose tolerance.
The study was funded by the UK MRC and Diabetes UK. Thanks to Sister Jean Gerrard and her staff for technical support, and to Drs Murray Stewart and David Humphriss for their contributions to the baseline studies.