Increased prevalence of diabetes among immigrants from non-European countries in 60-year-old men and women in Sweden.

AIMS: To estimate the prevalence of risk factors for diabetes in subjects of foreign origin compared to Swedish-born. METHODS: A cross-sectional study with subjects included from a random sample of the population of in Stockholm County, Sweden, among 60-year-old men and women. Number in the analyzed groups were Swedish-born n=3329, immigrants from Europe n=654, and immigrants from non-European countries n=123. The health screening including physical examination, laboratory testing and a questionnaire on medical, socioeconomic and life-style factors. RESULTS: Men showed a higher prevalence of diabetes than women, 9.7% vs. 5.1% (P<0.001). Among European immigrants, no excess risk was present compared to Swedish-born subjects, 7.6% vs. 6.9%, sex-adjusted odds ratio (OR): 1.14 (95% confidence interval (CI) 0.83-1.57). Among non-European immigrants, prevalence of diabetes was two times higher, i.e. 14.6%, sex-adjusted OR: 2.19 (95% CI: 1.30-3.69), compared to Swedish-born subjects. When also adjusting for anthropometrical, socioeconomic and life-style factors, ORs were for European immigrants 0.95 (95% CI: 0.68-1.33), and non-European immigrants OR: 1.21 (95% CI: 0.65-2.26). Beside sex, i.e. female OR: 0.50 (95% CI: 0.38-0.66), the only significant covariate in full model was BMI-category, i.e. overweight OR: 1.83 (95% CI: 1.29-2.59), and obesity OR: 4.64 (95% CI: 3.25-6.63). CONCLUSIONS: The prevalence of diabetes was found to be considerably higher among immigrants from non-European countries in Sweden. The excess risk was associated with above all to a higher BMI.

Glucose potentiation of arginine-induced insulin secretion is impaired in subjects with a glucokinase Glu256Lys mutation.

Insulin and glucagon release and insulin sensitivity were investigated in patients with glucokinase deficiency. Five subjects with a missense mutation (Glu256Lys) were studied. They were compared with six healthy subjects with low insulin response but normal glucose tolerance. Insulin and glucagon levels were measured at blood glucose 7.1 +/- 0.1 mmol/l and at 10.9 +/- 0.2 mmol/l with or without arginine (5 g i.v.). Insulin sensitivity was assessed as the ratio between infused glucose and the insulin level (M:I) during hyperglycemic clamps. Glu256Lys subjects were nonobese and had fasting blood glucose 6.7 +/- 0.1 mmol/l (P < 0.001 vs. control group). Insulin release was reduced in response to 11 mmol/l glucose (61% of control group, P < 0.05) as well as to arginine in the presence of 11 mmol/l glucose (54% of control group, P < 0.01). Also, the slope of potentiation, i.e., the enhancement of arginine-induced release as a function of prevailing glucose concentration, was reduced (delta insulin/delta glucose, 47% of control group, P < 0.05). As for glucagon release, the response to arginine was not inhibited normally by glucose, resulting in threefold higher levels at 11 mmol/l glucose versus control subjects. Insulin sensitivity, assessed as M:I, was significantly (P < 0.05) reduced (55% of control group). Glucokinase deficiency thus affects not only insulin responses to glucose per se but also glucose potentiation of responses to non-nutrient secretagogues. Abnormalities in glucagon release and insulin sensitivity coexist with attenuated insulin responses in glucokinase-deficient subjects.

Minimal increases in glucagon levels enhance glucose production in man with partial hypoinsulinemia.

In man a small dose of somatostatin (50 micrograms/h) suppressed moderately basal insulin (5 microU/ml) and glucagon (40 pg/ml) levels. This resulted in a short-lasting hypoglycemia, which was then followed by marginal hyperglycemia throughout the experiment. The addition of a minimal dose of glucagon (0.50 ng/kg/min) to somatostatin normalized basal glucagon levels and resulted in a significant and sustained hyperglycemia. During the first 2 h, hyperglycemia was mainly due to increased glucose production, whereas later on it was maintained by decreased glucose uptake. We conclude that, in man moderately deprived of insulin, even a marginal change in glucagon level induces a long-lasting hyperglycemia.

Glucose tolerance, insulin release, and insulin sensitivity in normal-weight women with previous gestational diabetes mellitus.

Out of 57 women with previous histories of gestational diabetes (GD), 23 were of normal weight postpartum and willing to participate in three studies characterizing oral glucose tolerance (OGTT), insulin responsiveness to intravenous glucose (glucose infusion test, GIT), and insulin sensitivity (somatostatin, insulin, and glucose infusion test, SIGIT). The experiments were performed 6-36 mo after cessation of breast-feeding. The control group comprised 10 healthy women with normal OGTT matched for age and weight. Among subjects with previous histories of GD, 9 had normal, 8 borderline, and 6 decreased OGTT. As a group, women with previous histories of GD have significantly decreased insulin response and insulin sensitivity. Furthermore, all 14 with borderline and decreased OGTT demonstrated a low early insulin response during GIT (5-min value below the upper border of the lower quartile of normals), whereas insulin sensitivity was normal in 6 and low in 8 (glucose values attained during SIGIT were lower or higher, respectively, than the lower border of the upper quartile of controls). The women with previous histories of GD and normal OGTT exhibited normal (n = 4) and low (n = 5) insulin responses. Three of the former subjects had low and the remaining 6 had normal insulin sensitivity. In conclusion, as many as 60% of normal-weight women with previous histories of GD had borderline or decreased OGTT 6-36 mo postpartum. This derangement could be due to impaired early insulin response, which in some subjects was combined with low insulin sensitivity. Follow-up of women with previous histories of GD might enlighten the pathogenesis of non-insulin-dependent diabetes mellitus.

Contributions by kidney and liver to glucose production in the postabsorptive state and after 60 h of fasting.

Contributions of renal glucose production to whole-body glucose turnover were determined in healthy individuals by using the arteriovenous balance technique across the kidneys and the splanchnic area combined with intravenous infusion of [U-13C6]glucose, [3-(3)H]glucose, or [6-(3)H]glucose. In the postabsorptive state, the rate of glucose appearance was 11.5 +/- 0.6 micromol x kg(-1) x min(-1). Hepatic glucose production, calculated as the sum of net glucose output (9.8 +/- 0.8 micromol x kg(-1) x min(-1)) and splanchnic glucose uptake (2.2 +/- 0.3 micromol x kg(-1) x min(-1)) accounted for the entire rate of glucose appearance. There was no net exchange of glucose across the kidney and no significant renal extraction of labeled glucose. The renal contribution to total glucose production calculated from the arterial, hepatic, and renal venous 13C-enrichments (glucose M+6) was 5 +/- 2%. In the 60-h fasted state, the rate of glucose appearance was 8.2 +/- 0.3 micromol x kg(-1) x min(-1). Hepatic glucose production, estimated as net splanchnic output (5.8 +/- 0.7 micromol x kg(-1) x min(-1)) plus splanchnic uptake (0.6 +/- 0.3 micromol x kg(-1) x min(-1)) accounted for 79% of the rate of glucose appearance. There was a significant net renal output of glucose (0.9 +/- 0.3 micromol x kg(-1) x min(-1)), but no significant extraction of labeled glucose across the kidney. The renal contribution to whole-body glucose turnover calculated from the 13C-enrichments was 24 +/- 3%. We concluded that 1) glucose production by the human kidney in the postabsorptive state, in contrast to recent reports, makes at most only a minor contribution (approximately 5%) to blood glucose homeostasis, but that 2) after 60-h of fasting, renal glucose production may account for 20-25% of whole-body glucose turnover.

Glucose turnover in hyperthyroid patients with normal glucose tolerance.

To determine the diabetogenic effect(s) of thyroid hormones, we simultaneously measured glucose turnover in six hyperthyroid patients and six normal subjects. All had normal fasting blood glucose concentration and oral glucose tolerance test values. We determined hepatic total glucose output (HTGO) and total glucose phosphorylation with [2-3H]glucose and hepatic glucose production (HGP) and irreversible glucose uptake using [6-3H]glucose. The difference between the two turnover rates indicates the extent of hepatic glucose cycling (glucose in equilibrium glucose-6-phosphate). Measurements were made both in the postabsorptive steady state and during a 2-h glucose infusion (11.1 mumol/kg.min). The postabsorptive HTGO and total glucose phosphorylation were increased in the hyperthyroid patients [13.5 +/- 0.8 (+/- SE) vs. 11.3 +/- 0.4 mumol/kg.min; P less than 0.05]. HGP and irreversible glucose uptake also were slightly but not significantly higher. During the glucose infusion, HTGO and HGP were less suppressed in the hyperthyroid patients than in the normal subjects, while the increments in peripheral glucose uptake were normal. In hyperthyroidism, glucose cycling was increased both postabsorptively (2.35 +/- 0.27 vs. 1.17 +/- 0.25 mumol/kg.min; P less than 0.025) and during glucose infusion (2.57 +/- 0.34 vs. 1.31 +/- 0.35 mumol/kg.min; P less than 0.05). We conclude that increases in HTGO and HGP are important features of hyperthyroidism, especially during glucose infusion. The increase in GC indicates increased activities of both glucokinase and glucose-6-phosphatase. The diabetogenic effect of hyperthyroidism, as revealed most markedly by [2-3H]glucose, could be accounted for by augmented glucose production, possibly due to increased glucose-6-phosphatase activity.

Quantitation of glycogen/glucose-1-P cycling in liver.

A method is introduced for quantitating cycling between hepatic glycogen and glucose-1-P in humans. It depends on the administration of trace [2-3H,6-14C]galactose, a glucose load, and acetaminophen. The ratio of 3H to 14C in the glucuronide of the acetaminophen excreted in urine to that in the administered galactose provides the measure of the fraction of glycogen synthesized that is synthesized from glucose-1-P formed from glycogen. The quantity of glucose-1-P formed from glycogen that is not reconverted to glycogen is not measured. It is assumed that the glucuronide samples the UDP-glucose pool in liver from which glycogen is formed, the last glucosyl units formed from UDP-glucose in glycogen synthesis are the first broken down, and the equilibration of [2-3H]glucose-1-P with fructose-6-P is rapid relative to its conversion to UDP-glucose. During a 5-hour period, while three normal subjects and three non-insulin-dependent diabetics, who had fasted overnight, were infused with 4 mg/kg/min of glucose, the rate of glycogen breakdown, as measured using the method, was only a small percentage of the rate of glycogen synthesis.

A probing dose of phenylacetate does not affect glucose production and gluconeogenesis in humans.

Phenylacetate ingestion has been used to probe Krebs cycle metabolism and to augment waste nitrogen excretion in urea cycle disorders. Phenylalkanoic acids, including phenylacetate, have been proposed as potential therapeutic agents in the treatment of diabetes. They inhibit gluconeogenesis in the liver in vitro and reduce the blood glucose concentration in diabetic rats. The effect of sodium phenylacetate ingestion on blood glucose and the contribution of gluconeogenesis to glucose production have now been studied in 7 type 2 diabetic patients. The study was not designed to test whether the changes in glucose metabolism observed in the rat could be reproduced in humans. After an overnight fast, over a period of 1 hour, 4.8 g phenylacetate was ingested, which is the highest dose used to probe Krebs cycle metabolism. Glucose production was measured by tracer kinetics using [6,6-(2)H2]glucose and gluconeogenesis by the labeling of the hydrogens of blood glucose on (2)H20 ingestion. The concentration of phenylacetate in plasma peaked by 2 hours after its ingestion, and about 40% of the dose was excreted in 5 hours. The plasma glucose concentration and production, and the contribution of gluconeogenesis to glucose production, were unaffected by phenylacetate ingestion at the highest dose used to probe Krebs cycle metabolism.

Quantitative contributions of gluconeogenesis to glucose production during fasting in type 2 diabetes mellitus.

Contributions of gluconeogenesis to glucose production were determined between 14 to 22 hours into a fast in type 2 diabetics (n = 9) and age-weight-matched controls (n = 7); ages, 60.4 +/- 2.3 versus 55.6 +/- 1.2 years and body mass indices (BMI) 28.6 +/- 2.3 versus 26.6 +/- 0.8 kg/m2. Production was measured using a primed-continuous [6,6-2H2]glucose infusion and gluconeogenesis from 2H enrichment at carbons 2 and 5 of blood glucose on 2H2O ingestion. Plasma glucose concentration declined from 9.6 +/- 0.6 at 14 hours to 7.3 +/- 0.6 at 22 hours in the diabetics (P = .001) and from 5.4 +/- 0.1 to 5.0 +/- 0.1 in the controls (P < .05). Production from the 17th to 22nd hour declined 27.1% +/- 0.6% in the diabetics versus 18.5% +/- 0.8% in the controls (P = .001); from 10.4 +/- 0.3 to 7.6 +/- 0.2 versus 10.0 +/- 0.4 to 8.2 +/- 0.4 micromol/kg/min. Percent contributions of gluconeogenesis to production measured at 1 1/2 to 2-hour intervals beginning the 15th hour were 6.8% +/- 1.0% more in the diabetics than controls. The quantity of glucose contributed by gluconeogenesis declined 19.8% +/- 3.8% (P < .001) in the diabetics and 6.9% +/- 2.3% in the controls (P = .05); 7.21 +/- 0.32 to 5.74 +/- 0.26 versus 6.20 +/- 0.28 to 5.75 +/- 0.24 micromol/kg/min. The contribution of glycogenolysis to production, estimated from the difference between production and gluconeogenesis, declined to the same extent in diabetic and control subjects, 40.7% +/- 6.6% and 37.7% +/- 4.1%; from 3.23 +/- 0.35 to 1.86 +/- 0.26 versus 3.81 +/- 0.22 to 2.42 +/- 0.28 micromol/kg/min. Thus, gluconeogenesis contributed more to glucose production in the diabetic than control subjects. Production and the contribution of gluconeogenesis declined more in the diabetic subjects during the fast. The factors regulating these changes remain uncertain.

Retinal microangiopathy and pigment epithelial lesions in subjects with normal, borderline, and decreased oral glucose tolerance.

Retinal fluorescein angiography was performed in 150 subjects: 64 with normal fasting blood glucose and normal oral glucose tolerance test (OGTT), 49 with borderline, and 37 with decreased OGTT. Microaneurysms were noted in only two subjects, both with decreased OGTT. Minute changes in the retinal pigment epithelium (RPE) were seen in 23% of the 64 normal persons, in 35% of those with borderline, and 49% of those with decreased OGTT (p less than 0.05). The impact of glucose intolerance was more pronounced in subjects under the age of 50 years, RPE changes being rare (7%) in those with normal OGTT but occurring in 32% of those with borderline or decreased OGTT (p less than 0.01). The corresponding figures among subjects aged 50 or more were 55% and 57%, respectively. We conclude that at least half of the subjects above 50 years show RPE alterations, and that minimal changes in glucose metabolism may precipitate the development of such changes at an earlier age.

Reoperation for primary hyperparathyroidism.

The reasons for failure of the initial exploration and the results of reoperation were analyzed in 53 patients with a diagnosis of primary hyperparathyroidism, 29 of whom were referred after initial operations elsewhere. Seventy-nine reoperations were performed. Sternotomy was used in 15 patients, and in retrospect was necessary in only 5. There was no operative mortality. The reasons for initial failure were incorrect diagnosis in 6 patients, true recurrence in 4 and persistent disease in 43. Persistence was caused by surgical failure in two thirds and pathology failure in one third. Of 47 patients reoperated on for hyperparathyroidism, 39 (83 percent) were cured, a rate warranting this type of surgery. Analysis of a long-term series of initial operations demonstrates a persistence rate of 5 percent (24 of 461) and a recurrence rate of 1 percent (4 of 461) in this disease. The need for reoperation was les than 1 percent over the recent 5 year period.

Glucose intolerance in uremic patients: the relative contributions of impaired beta-cell function and insulin resistance.

Glucose tolerance and tissue sensitivity to insulin were examined in 19 renal failure patients on chronic regular hemodialysis (group U) and in 6 matched control subjects with normal renal function (group A). Based on glucose tolerance as assessed by an oral glucose tolerance test (OGTT), glucose tolerance was normal in 5 (group U:N), borderline in 5 (group U:BL) and decreased in 9 uremic subjects (group U:D). Compared with group A the uremics demonstrated significantly (p less than 0.01) impaired insulin sensitivity as assessed by a continuous mixed infusion of somatostatin, insulin and glucose (SIGIT). In addition 19 non-diabetic subjects with normal fasting blood glucose and normal renal function, matching the uremic patients with respect to glucose tolerance as assessed by OGTT, were studied (group B). In group B impairments in both insulin secretion and insulin sensitivity tended to be more pronounced in subjects with decreased OGTT as compared with those with borderline OGTT. In contrast, insulin resistance was present to a similar degree in uremic subjects of group U:N, U:BL and U:D. During SIGIT endogenous insulin, glucagon and growth hormone (GH) were suppressed in both uremic and control subjects. This implies that insulin resistance in uremia is most likely not due to hyperglucagonemia or abnormal GH metabolism. During OGTT subjects of group U:N had significantly higher insulin response than subjects of group U:BL (p less than 0.02) and group U:D (p less than 0.01). Insulinogenic index was significantly higher in group U:N than in group U:BL (p less than 0.02) and group U:D (p = 0.01) and was higher in group U:BL than in group U:D (p less than 0.02).(ABSTRACT TRUNCATED AT 250 WORDS)

Relationship between insulin responses to D-glucose and to L-arginine in women with a history of gestational diabetes.

The relationship between insulin responses to glucose and to arginine was studied in non-obese women with previous gestational diabetes (PGD). One group, n = 10, had normal glucose tolerance (NGT) by WHO criteria and another, n = 8, had impaired glucose tolerance (IGT). A third group of women without PGD, n = 12, was also studied. A hyperglycaemic clamp (blood glucose level 11 mM) and an arginine stimulation test (150 mg/kg L-arginine followed by 10 mg/kg.min) were performed on separate days. The ratios of arginine to glucose responses 0-10 min differed: they were 1.00 for non-PGD, 1.29 for NGT and 1.46 for IGT (P < 0.02 vs non-PGD). A further difference between groups was the ratio between first- and second-phase glucose-induced insulin secretion, which was significantly decreased in IGT, 0.72, compared with NGT, 0.98 (P < 0.01), and non-PGD, 1.05 (P < 0.005). However, within each group insulin responses 0-10 min to glucose and arginine were strongly correlated: for NGT (r = 0.75, P < 0.05), for IGT (r = 0.85, P < 0.01) and for women without PGD (r = 0.69, P < 0.05). Insulin sensitivity, as assessed by the M/I ratio, was non-significantly decreased in IGT (0.18 +/- 0.03 mg/kg.min per mU/l vs 0.26 +/- 0.03 in NGT and 0.28 +/- 0.03 in non-PGD, P < 0.1).(ABSTRACT TRUNCATED AT 250 WORDS)

New probes to study insulin resistance in men; futile cycle and glucose turnover.

Insulin resistance has been measured in man by nonsteady state tracer methodology. Increase in overall glucose utilization and suppression of glucose production was measured when hyperglycemia was achieved either by infusing glucagon or glucose. With the first method, insulin resistance was assessed in obese man and in lean hypertriglyceridemic patients. With the second method, insulin resistance was assessed in lean mild type II diabetics. These methodologies can only assess deficiences in overall glucose utilization and glucose production, but cannot delineate the defect in glucose uptake by the liver. However, if a given metabolic event is essentially characteristic of only one organ, metabolic abnormalities specific to that organ can be detected in vivo provided there is a probe specific to that metabolic pathway. Therefore, in lean mild type II diabetics the liver glucose futile cycle was assessed by a double tracer method. Previously it was shown that liver glucose futile cycling is increased in diabetic dogs. In healthy control subjects in basal state and during glucose infusion, the futile cycle could not be detected, but it represented a major part of glucose metabolism in liver of type II diabetics. It appears, therefore, that most of the glucose taken up by the liver during the glucose challenge in diabetics reenters the blood stream without being oxidized or polymerized. On the basis of these studies, it was concluded that excessive hyperglycemia in the diabetics during glucose infusion is due to a decrease in irreversible glucose uptake (impaired phosphorylation and futile cycling) and to a decrease in suppression of glucose production. The relative contribution of the liver and periphery to hyperglycemia seems to be almost equivalent. The mechanism behind the increased glucose cycle activity is not clear. It may be due to a relative decrease of glycogen synthase or increase in glucose-6-phosphatase or both. These observations in mild lean type II diabetics may have implications also in some other types of diabetes, since we have observed that futile cycling is even more marked in obese type II diabetics and that it could account in part for the diabetogenic effect of growth hormone in acromegalics.

Psychosocial problems in patients with newly diagnosed diabetes: number and characteristics.

Early in the course of diabetes, it is important to identify and support patients whose psychosocial situations and reactions to the diagnosis may affect their ability to adjust or take adequate responsibility for self-care. We aimed to identify (a) the number and characteristics of patients, 18-65 years, newly diagnosed with diabetes, who needed psychosocial interventions and (b) the type of psychosocial problems they had. A total of 106 patients (72 men) were included in the study. Interviews showed that 41.5% had psychosocial problems. Fifteen dropped out early in the study; 38% of those remaining had psychosocial problems (PSP). More than half had problems with their life situation; most commonly in relationships. About a third had problems related to diabetes, most commonly, work-related. Compared to other participants, PSP patients lived in more strained social situations, especially regarding personal finances and social support. More of the PSP patients were anxious and depressed. They used negative coping strategies more often and more frequently expected that diabetes would negatively affect their future. In conclusion, early in the course of diabetes, screening instruments should be used to identify PSP patients. Treatment by medical social workers skilled in diabetes care should be offered.

Contribution of gluconeogenesis and glycogenolysis to hepatic glucose production in acromegaly before and after pituitary microsurgery.

The diabetogenic effect of excess growth hormone (GH) such as that in acromegaly is well known. However, the contribution of the various components to hepatic glucose production (HGP) is not completely understood. In this study we evaluated insulin resistance, HGP, gluconeogenesis (GNG), and glycogenolysis (GLY) in five patients with acromegaly before and after pituitary microsurgery. Insulin resistance was estimated by the HOMA index. HGP was measured using a primed continuous (6,6- 2H2) glucose infusion, and GNG was measured from 2 H enrichment at carbons 2 and 5 of blood glucose on ingestion of 2H2O. The ratio of these enrichments equals the fractional contribution of GNG to HGP, and GLY was calculated as the difference between HGP and GNG. All measurements were performed after 12 hours of fasting. Levels of GH and IGF-I decreased, as did the HOMA index (p<0.05). HGP was reduced from 11.4 micromol/kg/min to 9.7 micromol/kg/min (p=0.032). GNG contributed most to HGP. GNG was unchanged, whereas GLY's fraction decreased 29% (p=0.056) postoperatively. This pilot study indicates that GNG is the main contributor to HGP and that GLY is more sensitive than is GNG to the insulin resistance existing in acromegaly.

K-value and low insulin secretion in a non-obese white population: predicted glucose tolerance after 25 years.

AIMS/HYPOTHESIS: Insulin resistance and insulin deficiency are proposed as risk factors for IGT and type 2 diabetes. We assessed the predictive value of initial parameters for the outcome of an OGTT performed 24.3+/-2.9 years later in an unselected healthy non-obese population. METHODS: The K-value of an IVGTT was determined in 267 healthy subjects (mean+/-SD: age 31.0+/-12.0 years, BMI 21.8+/-2.8 kg/m(2)). First-phase insulin response to a glucose infusion test was estimated as an incremental 5- or 10-min (DeltaI5 or DeltaI10) value, and as insulinogenic indices (DeltaI5/DeltaG5 or DeltaI10/DeltaG10) adjusted for insulin sensitivity determined by homeostasis model assessment for insulin resistance ([DeltaI5/DeltaG5]/HOMA-IR). RESULTS: At follow-up, six subjects had type 2 diabetes and 47 had IGT; 214 retained normal glucose tolerance. Insulin sensitivity and early (30 min) insulin response decreased with decreasing outcome OGTT. Blood glucose (2 h) at OGTT correlated positively with initial age and BMI, and negatively with DeltaI5/DeltaG5, (DeltaI5/DeltaG5)/HOMA-IR and K-value. In multiple linear regression analysis, (DeltaI5/DeltaG5)/HOMA-IR, DeltaI10, K-value, age, HOMA estimate of insulin secretion, and fasting plasma glucose were significantly associated with 2-h OGTT blood glucose. Similar results were obtained on comparing differences between subjects with normal and decreased (IGT+diabetes) glucose tolerance. CONCLUSIONS/INTERPRETATION: In 267 non-obese healthy subjects, initial K-value and first-phase insulin response to glucose adjusted for insulin sensitivity, but not insulin sensitivity itself, were strong predictors of the outcome of an OGTT performed 25 years later. Thus, in contrast to obese or other high-risk populations, in lean subjects, decreased beta cell function, but not insulin resistance itself, determines future glucose tolerance.