Long-term repeatability of measures of early insulin secretion derived from an intravenous glucose tolerance test and conversion from impaired glucose tolerance to diabetes

Abstract

Aim. We assessed the long-term repeatability of the acute insulin response (AIR) and sensitivity index (S_I) derived from the frequently sampled intravenous glucose tolerance test (FSIGT).

Methods. An FSIGT was performed in 20 women who participated in a 6.5-month rye- and wheat-bread intervention trial, 70 men and women with impaired fasting glycaemia (IFG) or impaired glucose tolerance (IGT) who participated in the Genobin study, and 81 men and women with IGT who participated in the Finnish Diabetes Prevention Study (DPS).

Results. The correlation of AIR and S_I at base-line with respective values after the 6.5–8.5-month trials was 0.86–0.88 and 0.71–0.84, and before and after 4 years in the DPS substudy, 0.86 and 0.53. In multivariate analyses, AIR (relative risk for a 1-SD change, 0.67; 95% confidence intervals 0.46–0.97) predicted the conversion from IGT to diabetes in the DPS substudy.

Conclusion. AIR is highly repeatable even after 4 years of follow-up. The long-term repeatability of S_I is moderate. Our findings emphasize the importance of impaired early insulin secretion in the transition from IGT to diabetes, and the high degree of tracking of measures of early insulin secretion derived from the FSIGT.

• Impaired fasting glucose
• impaired glucose tolerance
• insulin secretion
• insulin sensitivity
• type 2 diabetes mellitus

Introduction

β-Cell dysfunction is a prerequisite for the development of impaired fasting glycaemia (IFG), impaired glucose tolerance (IGT), and type 2 diabetes, and is a primary determinant of which individuals with insulin resistance develop IFG, IGT, or eventually diabetes 1–10. Indeed, persons who progress from normal fasting and post-load glucose homeostasis to IFG or IGT and type 2 diabetes are characterized by a lower first-phase or acute insulin response (AIR) and further decline in the AIR with time. In contrast, non-progressors generally show little decline in β-cell function 4, 6–8, 10. Assessment of insulin secretion is therefore crucial to identify those at risk for further decline in glucose tolerance and development of diabetes, especially in high-risk individuals.

The minimal model analysis of the frequently sampled intravenous glucose tolerance test (FSIGT) is, along with the hyperglycaemic clamp, generally considered to be one of the gold standards for the assessment of insulin secretion in non-diabetic subjects 11, 12. Both impaired first-phase insulin secretion and insulin resistance as measured by the FSIGT have predicted development of diabetes 1, 4, 6, 7. Despite the importance of impaired insulin secretion as a determinant of worsening glucose tolerance and diabetes, little has been reported on the long-term repeatability of the AIR as measured during the FSIGT.

Key messages

• The acute insulin response measured during an intravenous glucose tolerance test is highly repeatable in individuals with normal or impaired glucose tolerance who do not develop diabetes, even after 4 years.

• Impaired first-phase insulin secretion plays a central role in the conversion from impaired glucose tolerance to type 2 diabetes.

Abbreviations

Table

AIR	acute insulin response
BMI	body mass index
DI	disposition index
DPS	Finnish Diabetes Prevention Study
FSIGT	frequently sampled intravenous glucose tolerance test
IFG	impaired fasting glycaemia
IGT	impaired glucose tolerance
ICC	intracorrelation coefficient
NCEP	National Cholesterol Education Program
OGTT	oral glucose tolerance test
QUICKI	quantitative insulin sensitivity check index
SI	insulin sensitivity index

We assessed the medium-term (2.5–8.5 months) and long-term (4 years) repeatability of the AIR measured during a FSIGT in relatively low-risk post-menopausal women participating in a rye- and wheat-bread intervention study 13, in men and women with the metabolic syndrome and IFG or IGT participating in the Genobin study, and in a subgroup of the Finnish Diabetes Prevention Study (DPS) 7. To provide further insight on the importance of early insulin secretion and insulin sensitivity in the pathophysiology of type 2 diabetes, we also compared measures of first-phase insulin secretion and insulin sensitivity derived from the FSIGT in the prediction of diabetes and worsening glucose tolerance in the subsample of the Finnish DPS.

Methods

Participants

The subjects were 20 post-menopausal women who participated in a rye- and wheat-bread intervention study 13, 70 men and women with the metabolic syndrome and IFG or IGT who participated in the Genobin study 14, and 81 men and women with IGT who participated in the DPS at the Kuopio centre 7, 15, 16 (Table I). An FSIGT was carried out at base-line and during the follow-up in all three studies to assess changes in insulin sensitivity and secretion. Subjects provided written informed consent. The Ethics Committee of the University of Kuopio and Kuopio University Hospital approved the studies.

Table I.  Characteristics at base-line. Values are presented as means (SD), proportions, or medians (interquartile ranges).

	Rye-/wheat-bread	Genobin	DPS substudy
n	20	70	81
Age (y)	59.2±6.0	60.2±6.5	55.5±7.2
Study design	Randomized cross-over	Randomized parallel	Randomized parallel
Control/intervention	20/20	16/54	37/44
Follow-up	2.5 and 6.5 months	8.5 months	4 years
Gender, male/female	0/20	36/34	31/50
Body mass index (kg/m^2)	27.5 (25.3, 29.7)	32.3 (30.9, 34.5)	30.8 (27.8, 34.0)
Fasting plasma glucose (mmol/L)	5.4 (5.0, 5.7)	6.4 (6.2, 6.8)	6.4 (5.7, 7.0)
2-h plasma glucose (mmol/L)	5.5 (4.4, 7.1)	7.3 (6.0, 8.8)	8.9 (8.1, 10.0)
Impaired fasting glycaemia	6/20	55/70	49/81
Impaired glucose tolerance	3/20	20/70	81/81
Early diabetes	0/20	14/70	0/81 ^a
Fasting serum insulin (mU/L)	9.1 (6.6, 12.8)	11.1 (7.6, 15.1)	15.0 (9.3, 20.5)
QUICKI	0.59 (0.54, 0.66)	0.44 (0.41, 0.48)	0.42 (0.39, 0.46)
FSIGT
SI	4.0 (3.2, 6.3)	2.0 (1.4, 2.7)	1.7 (0.9, 2.7)
AIR	392 (262, 578)	321 (205, 545)	143 (56, 255)
AIR (SI-adjusted)	389 (260, 579)	352 (217, 561)	134 (70, 279)
DI	1836 (1212, 2442)	649 (399, 1039)	224 (89, 530)

^a According to the 1985 World Health Organization criteria 18. At base-line, 14 men and women had fasting plasma glucose levels ranging from 7.0 to 7.7 mmol/L.

AIR = acute insulin response; DI = disposition index; DPS = Diabetes Prevention Study; FSIGT = frequently sampled intravenous glucose tolerance test; QUICKI = quantitative insulin-sensitivity check index; S_I=insulin sensitivity index; S_I-adjusted AIR = AIR adjusted by regression analysis for S_I.

For the post-menopausal women participating in the rye- and wheat-bread intervention study, the primary inclusion criteria were serum total cholesterol 5.0–8.5 mmol/L, non-high-density lipoprotein cholesterol 3.5–6.5 mmol/L, triacylglycerols < 2.5 mmol/L, and body mass index (BMI; body-weight in kg/height in m²) of 20–33 13. Only three women had IGT. A randomized cross-over study design with two 8-week bread periods was used.

For the Genobin study, 70 overweight or obese (BMI 28–40) men and women with IFG or IGT and two features of the metabolic syndrome (National Cholesterol Education Program (NCEP) criteria 17) participated. Fourteen subjects had early previously undiagnosed type 2 diabetes. Subjects were randomized to one of four groups: weight loss (n = 25), aerobic training (n = 15), resistance training (n = 14), or control (n = 18).

In the Finnish DPS, 522 subjects with IGT were randomized to an intervention or a control group in five centres as described elsewhere 15, 16. The main inclusion criteria were: BMI >25, age 40–64 years, and IGT according to the World Health Organization 1985 criteria 18. An oral glucose tolerance test (OGTT) was done at base-line and annually as described earlier in detail 15, 16. In the Kuopio centre, 81 men and women underwent a FSIGT at base-line.

Intervention measures

In the rye-bread intervention, the high-fibre rye- and white wheat-breads were intended to cover at least 20% of the daily intake of energy 13. Body-weight and insulin sensitivity as measured during the FSIGT did not change during the trial, but AIR improved during the rye-bread intervention 13.

In the Genobin study, the weight loss group underwent a 33-week weight loss and weight maintenance programme. The average weight loss was 4.6 kg. Insulin sensitivity as measured during the FSIGT tended to improve (P = 0.080). The aerobic training group increased training to ≥4 times/week for ≥30 min/session at 55%–65% of the maximum heart rate. The resistance training group increased training to 3–4 sessions/week at 70% one-repetition maximum, 1–2 sets×10–12 repetitions for upper and lower body muscle groups. Body composition, insulin sensitivity, and insulin secretion did not change during the trial in the exercise groups.

In the DPS, the intervention group received detailed advice about how to achieve the goals of the intervention, which were a reduction in weight of ≥5%, in the total intake of fat <30% of energy, and the intake of saturated fat <10% of energy consumed; an increase in fibre intake to ≥15 g/1,000 kcal; and moderate exercise for ≥30 min/day 15, 16. The control group received general advice. The intervention group lost about 3 kg in body-weight, with an improvement in insulin sensitivity during the FSIGT 7, 16.

Determination of IFG, IGT, and diabetes mellitus

For the 2-h OGTT, samples for glucose and insulin were taken before (fasting) and 2 h after a glucose load (75 g). For the rye-bread study and Genobin, IFG was defined as fasting plasma glucose 5.6–6.9 mmol/L, IGT as fasting plasma glucose <7.0 mmol/L and 2-hour plasma glucose 7.8–11.0 mmol/L, and type 2 diabetes as fasting glucose ≥7.0 mmol/L or 2-h glucose >11.0 mmol/L 19. In the DPS, IGT was defined according to the 1985 criteria of the World Health Organization 18 as fasting plasma glucose <7.8 mmol/L and 2-hour plasma glucose 7.8–11.0 mmol/L, and diabetes as fasting plasma glucose ≥7.8 mmol/L or 2-hour plasma glucose ≥11.1 mmol/L.

Frequently sampled intravenous glucose tolerance test

The FSIGTs were performed during the run-in period and at the end of rye- and wheat-bread periods, meaning that for each participant an FSIGT was carried out at base-line, at 2.5 months, and at 6.5 months 13. In the Genobin study FSIGTs were performed before and after the 33-week life-style intervention. In the DPS substudy, the FSIGT was carried out at base-line in 81 subjects and 4 years later in 52 subjects who remained non-diabetic during the 4-year follow-up 7.

The minimal model FSIGT was performed as described by Bergman 12 and in our earlier publication 7. An intravenous catheter was inserted into the antecubital veins of both arms. Glucose (330 mg/kg of body-weight) was given intravenously as a 50% solution in 1.5 min followed by 10 mL of 0.9% NaCl solution. Thereafter, a 0.9% NaCl solution was slowly infused. A bolus of 0.03 units/kg of insulin was injected 20 min after the glucose dose. The NaCl infusion was continued for 1.5 min after the insulin dose. Venous blood samples were collected for glucose and insulin determinations before the glucose dose and 23 times after the glucose dose (at 2, 4, 6, 8, 10, 12, 14, 16, 19, 22, 24, 27, 30, 40, 50, 60, 70, 90, 100, 120, 140, 160, and 180 min) via a catheter in the contralateral arm. To arterialize the venous blood, the arm was kept in a 50°C electric pad.

Plasma glucose was analysed by an enzymatic photometric assay (rye study: Granutest 100, Merck, Damstadt, Germany; Genobin: Thermo Clinical Labsystems, Vantaa, Finland; DPS substudy: Glucose Auto & Stat, Model GA-110, Daiichi, Kyoto, Japan) using a Kone Specific Clinical Analyser (Kone Ltd, Espoo, Finland) and serum insulin by radioimmunoassay (Phadaseph Insulin RIA 100, Pharmacia Diagnostica, Uppsala, Sweden) and, in the Genobin study, by a chemiluminescence sandwich method (ACS, Bayer A/S, USA).

The insulin sensitivity index (S_I) was calculated with the Minmod program 12. The AIR was calculated as the area under the insulin curve from 0 to 10 minutes. The relationship between insulin sensitivity and secretion has been described as hyperbolic (an inverse linear relationship when the log of insulin sensitivity and insulin secretion are plotted) 10, 20, 21. To take this into account, the disposition index (DI) was calculated as the product of AIR and S_I 22, 23. As an alternative method, we adjusted AIR by S_I, using regression analysis to adjust the log of AIR by the log of S_I and then transformed the S_I-adjusted AIR back into untransformed values by taking the antilog 24. The QUICKI insulin sensitivity index was calculated as (log(insulin concentration)) + log(glucose concentration))⁻¹ 25.

Statistical analysis

Data are displayed as means (SD) or as medians (interquartile ranges). Partial correlation analysis was used to assess the association of measures of insulin and glucose metabolism with corresponding measurements during the follow-up. The interventions themselves affect indices of glucose and insulin metabolism, for which reason partial correlation analysis (with adjustment for the intervention) was used instead of intracorrelation coefficient (ICC) analysis. To test for possible interactions with the interventions, we carried out univariate general linear models with the respective follow-up measure of insulin and glucose metabolism as the outcome variable, and an interaction term for the respective base-line measure and the intervention was also tested. No significant interaction of the interventions with the base-line measures of insulin and glucose metabolism was found. Cox proportional hazards regression analysis was used to assess the risk of incident diabetes associated with measures of the FSIGT. Univariate general linear models were used to assess the association of measures of the FSIGT with changes in the fasting and 2-h plasma glucose during the 4-year follow-up. P < 0.05 was considered statistically significant. SPSS for Windows v. 11 (Chicago, Illinois) was used.

Results

Base-line

All participants in the DPS substudy had IGT (1985 criteria), whereas only 3 of the 20 women in the rye-bread intervention had IGT (Table I). This difference was also reflected in the BMI and measures of insulin sensitivity and secretion. All participants of the Genobin study had IFG or IGT and the metabolic syndrome as defined by the NCEP, and more closely resembled the DPS participants with respect to BMI and measures of insulin and glucose metabolism. All participants in the rye-bread intervention were women, whereas 50 of 81 participants in the DPS substudy and 34 of 70 in the Genobin study were women.

Measures of insulin and glucose metabolism at base-line and follow-up and their partial correlations

In the combined groups of the rye- and wheat-bread study, the measures of insulin and glucose metabolism did not change during the follow-up (Table II). The group-adjusted partial correlation of base-line AIR with that measured 2.5 and 6.5 months later was 0.85–0.86 (Table II). The corresponding partial correlation for S_I-adjusted AIR was also high. Those of S_I and DI were lower (partial r = 0.66–0.71) and similar to those of fasting insulin and the QUICKI index.

Table II.  Measures of insulin and glucose metabolism in the rye-wheat intervention (n=20), Genobin study (n=70) and Finnish Diabetes Prevention Study (n=52).

Rye-wheat study	Base-line (0 months)	2.5 months	6.5 months	partial r (0–2.5 months)	partial r (0–6.5 months)
FSIGT
SI	4.0 (3.2, 6.3)	3.5 (2.5, 5.5)	3.4 (2.5, 5.1)	0.67	0.71
AIR	392 (262, 578)	383 (233, 619)	421 (227, 739)	0.85	0.86
AIR (SI-adjusted)	389 (260, 579)	381 (256, 592)	450 (242, 648)	0.83	0.78
DI	1816 (848)	1691 (1152)	1697 (1041)	0.70	0.66
Measures derived from fasting insulin and glucose concentrations
Plasma glucose (mmol/L)	5.4 (0.4)	5.3 (0.3)	5.5 (0.4)	0.64	0.57
Serum insulin (mU/L)	9.1 (6.6, 12.8)	6.9 (5.6, 8.8)	7.3 (6.1, 10.2)	0.67	0.65
QUICKI	0.59 (0.06)	0.64 (0.06)	0.61 (0.08)	0.67	0.70
Genobin study		Base-line (0 wk)	33 weeks		partial r
FSIGT
SI		2.0 (1.4, 2.7)	2.1 (1.6, 2.9)		0.83
AIR		321 (205, 545)	396 (214, 670)		0.88
AIR (SI-adjusted)		350 (217, 560)	411 (227, 618)		0.84
DI		648 (399, 1039)	833 (479, 1230) ^a		0.82
Fasting plasma glucose (mmol/L)		6.1 (5.7, 6.6)	6.1 (5.7, 6.8)		0.36
2-h glucose (mmol/L)		8.6 (7.7, 9.5)	8.4 (7.3, 10.4)		0.49
Serum insulin (mU/L)		11 (8, 15)	12 (7, 15)		0.67
QUICKI		0.44 (0.41, 0.48)	0.44 (0.42, 0.49)		0.69
Finnish DPS substudy		Base-line	4 years		partial r
FSIGT
SI		1.9 (1.0, 2.8)	2.2 (1.5, 3.0) ^b		0.53
AIR		187 (72, 365)	190 (96, 294)		0.89
AIR (SI-adjusted)		192 (74, 319)	204 (103, 300)		0.86
DI		312 (153, 616)	408 (180, 754)		0.72
Fasting plasma glucose (mmol/L)		6.1 (5.7, 6.6)	6.1 (5.7, 6.8)		0.55
2-h glucose (mmol/L)		8.6 (7.7, 9.5)	8.4 (7.3, 10.4)		0.00
Fasting serum insulin (mU/L)		13 (9, 17)	10 (8, 15) ^c		0.67
QUICKI		0.42 (0.40, 0.47)	0.45 (0.42, 0.48)		0.67

For the rye-wheat study, if r≥0.40, P<0.001. For Genobin and the DPS, if r=0.56–0.69, P=0.01–0.001; if r≥0.70, P<0.001. For the change from base-line to follow-up, paired samples t test: ^a P=0.017; ^b P<0.001; ^c P=0.036.

AIR = acute insulin response; DI = disposition index; DPS = Diabetes Prevention Study; FSIGT = frequently sampled intravenous glucose tolerance test; QUICKI = quantitative insulin-sensitivity check index; S_I=insulin sensitivity index; S_I-adjusted AIR = AIR adjusted by regression analysis for S_I.

In the combined groups of the Genobin study, only the DI changed significantly during follow-up (Table II). The group-adjusted partial correlation of base-line AIR, S_I-adjusted AIR, S_I, and DI with that measured 33 weeks later was 0.82–0.88. Those for fasting and post-load glucose concentrations were modest (partial r = 0.36–0.49).

In the combined groups of the DPS substudy, S_I increased and fasting insulin decreased during the 4-year follow-up (Table II). The difference between groups for the changes during follow-up in the variables in Table III were not significant 7. The age- and intervention group-adjusted partial correlation of AIR and S_I-adjusted AIR with that measured 4 years later was 0.86–0.89 (Table II). The corresponding partial correlations for S_I and DI were more modest (partial r = 0.53–0.72). The base-line 2-h glucose measurement was not associated with the follow-up measurement 4 years later. In subgroup partial correlation analyses, the repeatability of the measures of insulin and glucose metabolism described above was similar in the control and intervention groups (not shown).

Table III.  Relative risk (95% confidence intervals) of developing type 2 diabetes according to a 1-SD change in indices derived from the frequently sampled intravenous glucose tolerance test at base-line in 81 participants of the Finnish Diabetes Prevention Study.

FSIGT	Model 1 ^a	Model 2 ^b	Model 3 ^c	Model 4 ^d
SI	0.82 (0.51–1.30)	0.80 (0.50–1.26)	1.09 (0.67–1.80)	2.08 (1.00–4.95)
AIR	0.55 (0.42–0.73)	0.54 (0.41–0.72)	0.67 (0.46–0.97)	0.66 (0.45–0.98)
AIR (adjusted)	0.55 (0.42–0.73)	0.54 (0.41–0.72)	0.66 (0.46–0.97)	0.66 (0.45–0.98)
DI	0.38 (0.25–0.59)	0.38 (0.25–0.59)	0.53 (0.30–0.92)	0.49 (0.27–0.91)

^a Model 1: adjusted for age, sex and randomization group.

^b Model 2: adjusted for Model 1 and AIR (for S_I) or S_I (for AIR, adjusted AIR and DI).

^c Model 3: adjusted for Model 2 and base-line fasting and 2-hour glucose concentrations.

^d Model 4: adjusted for Model 3 and BMI.

AIR = acute insulin response; DI = disposition index; FSIGT = frequently sampled intravenous glucose tolerance test; S_I=insulin sensitivity index; S_I-adjusted AIR = AIR adjusted by regression analysis for S_I.

Intracorrelation coefficient

In analyses of the repeatability of measures derived from the FSIGT in the control group of the DPS (n = 21), the ICC was similar to the partial correlation coefficients shown in Table II (e.g., for AIR, ICC = 0.87; for S_I, ICC = 0.50).

FSIGT and prediction of diabetes

In the DPS substudy, 19 of 81 men and women developed diabetes. S_I did not predict diabetes, except for a border-line association in model 4 (Table III). In this model, S_I predisposed to diabetes but probably because of multicollinearity of the variables. AIR, S_I- adjusted AIR, and especially DI were inversely associated with incident diabetes, even after multivariate adjustment.

FSIGT and the change in fasting and 2-h plasma glucose levels

After adjustment for age, sex, randomization group, S_I, and base-line 2-h glucose concentrations, AIR, S_I- adjusted AIR and DI were inversely associated (β = − 0.29 to −0.31, P = 0.014–0.050) with the changes in fasting plasma glucose levels during the 4-year follow-up. Further adjustment for BMI slightly increased the association (β = − 0.36 to −0.38, P = 0.015–0.020). AIR, S_I-adjusted AIR, and DI were not associated with changes in 2-h glucose concentrations. S_I was not associated with changes in either fasting or 2-h glucose concentrations.

Discussion

AIR and S_I-adjusted AIR derived from the FSIGT are highly repeatable measures of early-phase insulin secretion even after up to 4 years of follow-up. The long-term repeatability of S_I, however, was less, comparable with that of fasting insulin levels or QUICKI. As a consequence, the DI, the product of S_I and AIR, was less repeatable than AIR and S_I-adjusted AIR. The DI nonetheless predicted incident type 2 diabetes during the 4-year follow-up in the DPS substudy better than the other indices of early insulin secretion, and better than S_I.

The relationship between early insulin secretion and insulin sensitivity has been described as rectangular hyperbolic, in which changes in insulin sensitivity are compensated for by changes in insulin secretion in the opposite direction 10, 20, 21. Because of this physiological relationship, insulin sensitivity must be taken into account when assessing first-phase insulin secretion. This has been most commonly done by calculating the DI 22, 23, which for the FSIGT is the product of AIR and S_I. The DI is not only a measure of insulin secretion, however, but also of insulin sensitivity, because S_I is a term in the equation. The hyperbolic relationship can be transformed mathematically into a linear inverse association by taking the log of AIR and the log of S_I. Insulin sensitivity can then be taken into account by adjustment of log(AIR) by log(S_I) using linear regression 24. The advantage of this method is that one obtains a measure of acute insulin secretion that is truly independent of S_I.

The repeatability of S_I-adjusted AIR was almost as high AIR without adjustment. The high repeatability of the AIR over a 1-week period (r=0.92) has been demonstrated 26. Our findings show that the high repeatability extends to the medium (2.5–8.5 months) and long (4 years) term. Even in IGT, early insulin secretion thus has a high degree of tracking over at least 4 years, at least in those who do not develop type 2 diabetes.

The high degree of tracking of AIR may partly be because early-phase insulin secretion is under tight physiological and genetic regulation 20, 21, 27–29. Postprandial glycaemic responses to a variety of mixed meals are maintained within a narrow range in glucose-tolerant individuals. First-phase insulin secretion is also normally physiologically up-regulated in response to obesity and increasing insulin resistance 20, 21. In prone individuals, however, β-cell failure occurs, and IGT or diabetes develops. The capacity for insulin secretion to compensate for increasing obesity and insulin resistance has a strong genetic component 28–30. Based on twin studies, about 55% of AIR and, based on more complicated modelling, possibly up to 76% of AIR is genetically determined 28, 29.

AIR derived from the FSIGT is, along with acute insulin secretion measured during the hyperglycaemic clamp, generally considered to be one of the gold standards for the assessment of insulin secretion 11, 12. These intravenous methods by-pass the gastro-intestinal component of insulin secretion, however. Incretins play a key role in mediating insulin secretion, accounting for up to two-thirds of the insulin secretion in response to a glucose load during the OGTT 31. Incretins may also be important in the pathogenesis of deteriorating glucose tolerance and type 2 diabetes 31. Therefore, the insulinogenic index or other measures derived from the OGTT or from mixed meal feeding studies may be more physiological if less precise in the measurement of early insulin secretion. Both FSIGT and OGTT measures of early insulin secretion have predicted incident type 2 diabetes in prospective studies 1–10.

The repeatability of the DI was lower in the rye-wheat intervention and DPS substudy. The much lower repeatability of DI was because of the lower repeatability of S_I implicit in its calculation. The low repeatability of S_I in the rye-wheat study may have been because of the rye-wheat intervention or the small size of the study. The low repeatability of S_I is not due to the intervention in the DPS, however, because the repeatability of S_I was similarly low in the control group. The low 4-year repeatability of S_I may therefore be a reflection of actual variability in peripheral insulin resistance over time.

Measures of impaired early insulin secretion derived from the FSIGT strongly predicted conversion from IGT to type 2 diabetes and adverse changes in fasting glucose concentrations during the 4-year follow-up in the DPS substudy. In contrast, insulin sensitivity as measured by S_I did not predict diabetes or changes in glucose concentrations. In those who did not develop diabetes, however, insulin sensitivity improved by 34% during the 4-year follow-up (and tended to improve more in the intervention group 7), whereas measures of insulin secretion did not significantly change.

Insulin sensitivity as measured by S_I seemed to actually increase the risk of diabetes in the multivariate model including BMI, but this is likely due to multicollinearity. The reason why S_I was not a determinant of diabetes in the DPS subsample may be because of the small sample size and because the DPS was limited to overweight individuals with IGT. In the much larger US Diabetes Prevention Program study (DPP), insulin resistance and early insulin secretion as estimated crudely based on fasting and OGTT-derived insulin and glucose concentrations also predicted development of diabetes during the trial 9. In large cohort studies, both insulin sensitivity and insulin secretion as measured by the FSIGT have predicted incident diabetes 8, 10. Our findings nonetheless underscore the importance of impaired insulin secretion in the progression from IGT to type 2 diabetes 7.

None of the FSIGT indices predicted adverse changes in 2-h glucose concentrations in the DPS. The 2-h glucose values at base-line and the 4-year follow-up were not correlated at all. Two-hour glucose concentrations during an OGTT have a low repeatability 32, 33. Moreover, the range of values was restricted to IGT, and the follow-up was long.

Prevention of type 2 diabetes by life-style changes in individuals with IGT probably occurs mainly through improvement of insulin sensitivity rather than through direct effects on insulin secretion 7, 34, 35. Nonetheless, certain dietary changes 13, 24, 36, 37, increased physical activity 38, and weight loss 39 may also directly improve early-phase insulin secretion. Recent findings from the DPP with measures of insulin sensitivity and secretion based on fasting and post-load insulin and glucose values also suggest that life-style changes can enhance β-cell function independently of improvements in insulin resistance 9.

A limitation of this study is that all of the women in the rye-wheat study, most of the men and women in the Genobin study, and about half of the participants in the DPS substudy underwent life-style interventions that could alter insulin secretion or insulin sensitivity. Nonetheless, the results were similar in the control arm of the DPS. Moreover, in all of these studies, spanning in duration from 2.5 months up to 4 years, of subjects ranging from mainly normoglycaemic to IGT, AIR was highly repeatable, again underscoring that AIR is to a large degree determined by non-environmental factors.

AIR and S_I-adjusted AIR derived from the FSIGT are highly repeatable measures of early-phase insulin secretion even after up to 4 years of follow-up. The repeatability of S_I was at best modest. DI was less repeatable than AIR and S_I-adjusted AIR, but was a stronger predictor of incident type 2 diabetes. Caution should be taken when considering DI as an index of early insulin secretion, however, because S_I is implicit in its calculation. Our findings, based on three independent study groups, emphasize the high degree of tracking of measures of early insulin secretion derived from the FSIGT in men and women with IGT.

Acknowledgements

This work was supported by grants from the Academy of Finland (40758 to M.U., 46558 to J.T., 209445 M.K. and 104943 to R.R.), the EVO-fund of the Kuopio University Hospital (no. 5106 to M.U.), the Finnish Diabetes Foundation, the Ministry of Education of Finland, the Diabetes Research foundation, and the Technology Development Center of Finland. Fazer Bakeries Ltd and Vaasan & Vaasan Oy provided breads for the rye-wheat intervention. We also thank the staff of the Department of Clinical Nutrition, University of Kuopio, for their skill in carrying out the intravenous glucose tolerance tests and other work in this study. None of the authors have competing interests related to this study.