Development of Pericardial Fat Count Images Using a Combination of Three Different Deep-Learning Models: Image Translation Model From Chest Radiograph Image to Projection Image of Three-Dimensional Computed Tomography

RATIONALE AND OBJECTIVES

Pericardial fat (PF)-the thoracic visceral fat surrounding the heart-promotes the development of coronary artery disease by inducing inflammation of the coronary arteries. To evaluate PF, we generated pericardial fat count images (PFCIs) from chest radiographs (CXRs) using a dedicated deep-learning model.

MATERIALS AND METHODS

We reviewed data of 269 consecutive patients who underwent coronary computed tomography (CT). We excluded patients with metal implants, pleural effusion, history of thoracic surgery, or malignancy. Thus, the data of 191 patients were used. We generated PFCIs from the projection of three-dimensional CT images, wherein fat accumulation was represented by a high pixel value. Three different deep-learning models, including CycleGAN were combined in the proposed method to generate PFCIs from CXRs. A single CycleGAN-based model was used to generate PFCIs from CXRs for comparison with the proposed method. To evaluate the image quality of the generated PFCIs, structural similarity index measure (SSIM), mean squared error (MSE), and mean absolute error (MAE) of (i) the PFCI generated using the proposed method and (ii) the PFCI generated using the single model were compared.

RESULTS

The mean SSIM, MSE, and MAE were 8.56 × 10-1, 1.28 × 10-2, and 3.57 × 10-2, respectively, for the proposed model, and 7.62 × 10-1, 1.98 × 10-2, and 5.04 × 10-2, respectively, for the single CycleGAN-based model.

CONCLUSION

PFCIs generated from CXRs with the proposed model showed better performance than those generated with the single model. The evaluation of PF without CT may be possible using the proposed method.

The efficacy of 18F-FDG-PET-based radiomic and deep-learning features using a machine-learning approach to predict the pathological risk subtypes of thymic epithelial tumors

OBJECTIVE

To examine whether the machine-learning approach using 18-fludeoxyglucose positron emission tomography (18F-FDG-PET)-based radiomic and deep-learning features is useful for predicting the pathological risk subtypes of thymic epithelial tumors (TETs).

METHODS

This retrospective study included 79 TET [27 low-risk thymomas (types A, AB and B1), 31 high-risk thymomas (types B2 and B3) and 21 thymic carcinomas] patients who underwent pre-therapeutic 18F-FDG-PET/CT. High-risk TETs (high-risk thymomas and thymic carcinomas) were 52 patients. The 107 PET-based radiomic features, including SUV-related parameters [maximum SUV (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG)] and 1024 deep-learning features extracted from the convolutional neural network were used to predict the pathological risk subtypes of TETs using six different machine-learning algorithms. The area under the curves (AUCs) were calculated to compare the predictive performances.

RESULTS

SUV-related parameters yielded the following AUCs for predicting thymic carcinomas: SUVmax 0.713, MTV 0.442, and TLG 0.479 or high-risk TETs: SUVmax 0.673, MTV 0.533, and TLG 0.539. The best-performing algorithm was the logistic regression model for predicting thymic carcinomas (AUC 0.900, accuracy 81.0%), and the random forest (RF) model for high-risk TETs (AUC 0.744, accuracy 72.2%). The AUC was significantly higher in the logistic regression model than three SUV-related parameters for predicting thymic carcinomas, and in the RF model than MTV and TLG for predicting high-risk TETs (each; p < 0.05).

CONCLUSION

18F-FDG-PET-based radiomic analysis using a machine-learning approach may be useful for predicting the pathological risk subtypes of TETs.

ADVANCES IN KNOWLEDGE

Machine-learning approach using 18F-FDG-PET-based radiomic features has the potential to predict the pathological risk subtypes of TETs.

P2716Correlation between computed tomography derived ischemia index and conventional fractional flow reserve

Background

Computed tomography (CT) derived ischemia index is a novel tool to determine the significance of coronary artery disease, however, its ability to identify myocardial ischemia has not been examined.

Methods

From Jan. 2013 to Dec. 2015, 14,000 patients underwent ECG-gated CT coronary angiography for suspected coronary artery disease. From Jan. 2013 to Jan. 2016, 483 patients (589 vessels) underwent conventional FFR to assess myocardial ischemia. Among those, FFR was assessed in 148 patients (167 vessels) within 30 days after CT coronary angiography. We further excluded 24 patients with prior stenting or bypass grafting, 3 patients with multiple MDCT or FFR, and 6 patients with insufficient datasets. Finally, we included 117 patients (127 vessels) to assess the correlation between CT derived ischemia index and conventional FFR.

CT derived ischemia index was calculated as follows: left ventricular volume distally to the coronary artery lumen (cm3)/coronary artery lumen area (mm2). Left ventricular volume was automatically determined using Advantage Workstation and divided according to the nearest coronary artery tree. Center of coronary arteries were manually traced and contours of coronary artery lumen were automatically depicted. CT derived ischemia index was calculated at approximately every 0.625mm point of coronary artery. Moving median of consecutive 5 points (approximately 3.125mm) were used to remove outliers. Maximum value of CT derived ischemia index per coronary artery segment was calculated to determine the significance of coronary artery disease.

Results

Mean age was 71.3±10.5 years and 63.8% of patients were male. Coronary angiography was performed to assess conventional FFR at the median of 13 days (IQR 7 to 18 days) after CT. Majority of the target vessel was left anterior descending artery (71.7%), followed by right coronary artery (14.2%), left circumflex artery (13.4%) and left main coronary artery (0.8%). According to the quantitative coronary angiography, minimum lumen diameter was 1.47±0.32mm with percent diameter stenosis of 48.3±10.4%. Median FFR value was 0.83 (IQR 0.76 to 0.88) and positive test for myocardial ischemia (FFR <0.80) was observed in 42 vessels (33.1%). Maximum CT derived ischemia index per segment ranged from 1.825 to 57.296 (median 8.333, IQR 4.911 to 14.484). There was a negative correlation between CT derived ischemia index and FFR (r=−0.319, 95% confidence interval −0.467 to −0.153, P<0.001). Receiver operating characteristic analysis indicated CT derived ischemia index of 9.962 has 76.2% sensitivity and 70.6% specificity for the presence of FFR<0.80 (AUC 0.73, 95% CI 0.64 to 0.82).

Conclusions

A novel tool of CT derived ischemia index has a significant negative correlation with conventional FFR in lesions with mild to moderate stenosis. Larger multicenter prospective studies are needed to fully determine the impact of CT derived ischemia index.

PCLO rs2522833-mediated gray matter volume reduction in patients with drug-naive, first-episode major depressive disorder

Major depressive disorder (MDD) has been linked to differences in the volume of certain areas of the brain and to variants in the piccolo presynaptic cytomatrix protein (PCLO), but the relationship between PCLO and brain morphology has not been studied. A single-nucleotide polymorphism (SNP) in PCLO, rs2522833, is thought to affect protein stability and the activity of the hypothalamic-pituitary-adrenal axis. We investigated the relationship between cortical volume and this SNP in first-episode, drug-naive patients with MDD or healthy control subjects. Seventy-eight participants, including 30 patients with MDD and 48 healthy control subjects, were recruited via interview. PCLO rs2522833 genotyping and plasma cortisol assays were performed, and gray matter volume was estimated using structural magnetic resonance images. Among the individuals carrying the C-allele of PCLO rs2522833, the volume of the left temporal pole was significantly smaller in those with MDD than in healthy controls (family-wise error-corrected, P=0.003). No differences were detected in other brain regions. In addition, the C-carriers showed a larger volume reduction in the left temporal pole than those in the individuals with A/A genotype (P=0.0099). Plasma cortisol levels were significantly higher in MDD-affected C-carriers than in the healthy control C-carriers (12.76±6.10 vs 9.31±3.60 nm, P=0.045). We conclude that PCLO SNP rs2522833 is associated with a gray matter volume reduction in the left temporal pole in drug-naive, first-episode patients with MDD carrying the C-allele.

Inverse correlation between activated protein C generation and carotid atherosclerosis in Type 2 diabetic patients

AIMS

Activated protein C (APC) is a key regulator of the clotting system and immune responses. We studied the relationship between the degree of atherosclerosis as measured by the intima-media thickness (IMT) of carotid artery and APC generation in Type 2 diabetic patients.

METHODS

Eighty-seven Type 2 diabetic patients and 35 control subjects participated. APC generation was assessed by the plasma APC-protein C inhibitor complex (APC-PCI) levels and the mean IMT of carotid artery was measured by ultrasonography. The plasma levels of the thrombin-anti-thromobin complex (TAT) and platelet-derived growth factor (PDGF) were measured by enzyme-linked immunoassays.

RESULTS

Plasma TAT levels were significantly higher in diabetic patients [2.03 (1.12, 2.56) ng/ml, median (25th, 75th percentile)] compared with control subjects [0.85 (0.55, 2.08) ng/ml, P < 0.01]. Plasma APC-PCI levels were significantly lower in diabetic patients [0.93 (0.74, 1.22) ng/ml], than in control subjects [1.66 (1.25, 2.36) ng/ml, P < 0.001]. The mean IMT was significantly increased in diabetic patients (0.881 +/- 0.242 mm; mean +/- sd) compared with control subjects (0.669 +/- 0.140 mm; P < 0.01). Univariate analysis showed a significant and inverse correlation between plasma APC-PCI levels and mean IMT (r = -0.32, P < 0.005), and multivariate regression analysis confirmed the independent correlation (P < 0.05). Moreover, plasma APC-PCI levels significantly and inversely correlated with plasma PDGF levels in diabetic patients (r = -0.30, P < 0.01).

CONCLUSIONS

These results suggest that decreased APC generation is associated with vascular atherosclerotic changes in Type 2 diabetic patients.

Plasma levels of agouti-related protein are increased in obese men

To investigate the relationship between peripheral blood levels of agouti-related protein (AGRP) and various parameters of obesity, we measured the plasma level of AGRP in 15 obese and 15 nonobese men and evaluated its relationship with body mass index (BMI), body fat weight, and visceral, sc, and total fat areas measured by computed tomography, fasting insulin levels, glucose infusion rate during an euglycemic hyperinsulinemic clamp study, serum leptin, and plasma alpha-MSH. Obese men had significantly higher plasma concentrations of AGRP than nonobese men (P < 0.01). Univariate analysis showed that the plasma levels of AGRP are proportionally correlated with BMI, body fat weight, and sc fat area in obese men (BMI: r = 0.732, P < 0.01; body fat weight: r = 0.603, P < 0.02; sc fat area: r = 0.668, P < 0.01) and in all men (BMI: r = 0.839, P < 0.0001; body fat weight: r = 0.818, P < 0.0001; sc fat area: r = 0.728, P < 0.0001). In all men, the plasma levels of AGRP were significantly correlated with the visceral fat area (r = 0.478, P < 0.01), total fat area (r = 0.655, P < 0.0001), fasting insulin level (r = 0.488, P < 0.01), glucose infusion rate (r = -0.564, P < 0.01), serum level of leptin (r = 0.661, P < 0.0001), and the plasma level of alpha-MSH (r = 0.556, P < 0.01). In all subjects, multiple regression analysis showed that the plasma levels of AGRP are significantly (F = 15.522, r = 0.801, P < 0.03) correlated with the plasma levels of alpha-MSH, independently from the total fat area. However, the correlation between plasma levels of AGRP and serum levels of leptin was found to be dependent on the total fat area. In brief, these findings showed that the circulating levels of AGRP are increased in obese men and that they are correlated with various parameters of obesity. Although correlation does not prove causation, the results of this study suggest that peripheral AGRP may play a role in the pathogenesis of obesity.

Homeostasis model assessment is a reliable indicator of insulin resistance during follow-up of patients with type 2 diabetes

OBJECTIVE

To investigate the usefulness of the homeostasis model assessment as an index of insulin resistance (HOMA-IR) for evaluating the clinical course of patients with type 2 diabetes.

RESEARCH DESIGN AND METHODS

The usefulness of HOMA-IR and its relationship with insulin resistance assessed by the hyperinsulinemic-euglycemic clamp study (clamp IR) were evaluated in 55 Japanese patients with type 2 diabetes before and after treatment. The patients were subjected to diet (approximately 1,440-1,720 kcal/day) and exercise therapy (walking 10,000 steps daily) for 6 weeks during their hospitalization.

RESULTS

Univariate regression analysis disclosed a significant correlation between log-transformed HOMA-IR and log-transformed clamp IR before (r = -0.613, P < 0.0001) and after ( = -0.734, P < 0.0001) treatment. Neither the slopes (-0.71 +/- 0.12 vs. -0.79 +/- 0.09, F = 0.25, P = 0.61) nor the intercepts (y-intercept = 1.67 vs. 1.70, x-intercept = 2.36 vs. 2.15, F = 0.02, P = 0.88) of the regression lines between HOMA-IR and clamp IR were significantly different before and after treatment. There was a significant correlation between the decrease in log-transformed HOMA-IR and the increase in clamp IR during treatment (r = -0.617, P < 0.0001).

CONCLUSIONS

HOMA-IR may constitute a useful method not only for diagnosing insulin resistance, but also for follow-up during the treatment of patients with type 2 diabetes.

A case of obesity, diabetes and hypertension treated with very low calorie diet (VLCD) followed by successful pregnancy with intrauterine insemination (IUI)

The patient was a 32-year-old obese woman with a history of type 2 diabetes and hypertension for 6 years. Although she was treated with antihypertensive agents and intensive insulin therapy, her hyperglycemia was difficult to control. She wanted to have a baby but pregnancy was not recommended because her diabetes was under poor control and the use of antihypertensive medication. To achieve good control of obesity, diabetes and hypertension, she was admitted to our clinical department for weight reduction using very low calorie diet (VLCD). During VLCD she had a 19.8 kg reduction in body weight and her blood glucose and blood pressure were in good control without the use of drugs. Five months later, she became pregnant after the fourth trial of intrauterine insemination (IUI) and gave birth to a female baby under insulin therapy. This is the first report that showed the usefulness of VLCD for prepregnant control of glucose metabolism and blood pressure in an obese hypertensive patient with type 2 diabetes mellitus.

Elevated plasma levels of alpha-melanocyte stimulating hormone (alpha-MSH) are correlated with insulin resistance in obese men

OBJECTIVE

The role of alpha-melanocyte stimulating hormone (MSH) in obesity has been well-documented. However, circulating alpha-MSH concentrations in obese men and their relationship with clinical indicators of obesity and glucose metabolism have not as yet been evaluated.

METHODS

We measured the plasma concentrations of alpha-MSH in 15 obese and 15 non-obese male subjects. The relationship of the plasma concentrations of alpha-MSH with body mass index (BMI), body fat mass (measured by bioelectric impedance), body fat distribution (measured by computed tomography), insulin levels, insulin resistance (assessed by the glucose infusion rate (GIR) during an euglycemic hyperinsulinemic clamp study) and with the serum concentrations of leptin and TNF-alpha were also evaluated.

RESULTS

In obese men, the plasma alpha-MSH concentrations were significantly increased compared with those in non-obese men (P< 0.02). The plasma levels of alpha-MSH were positively correlated with BMI (r= 0.560, P< 0.05), fasting insulin levels (r=0.528, P< 0.05) and with visceral fat area (r=0.716, P<0.01), but negatively correlated with GIR (r= -0.625, P< 0.02) in obese male subjects. There were significant correlations between plasma concentrations of alpha-MSH and visceral fat area (r=0.631, P< 0.02), and GIR (r = -0.549, P< 0.05) in non-obese male subjects. Circulating concentrations of alpha-MSH were not significantly correlated with the serum concentrations of leptin and TNF-alpha in both obese and non-obese men.

CONCLUSION

Circulating concentrations of alpha-MSH are significantly increased and correlated with insulin resistance in obese men.

Insulin sensitivity is not affected by mutation of codon 972 of the human IRS-1 gene

We investigated the relationship of codon 972 polymorphism of the insulin receptor substrate-1 (IRS-1) gene with insulin resistance in the Japanese population. Among 130 patients with type-2 diabetes mellitus (DM), we identified 6 who were heterozygous for the Gly972Arg mutation. Among 144 healthy subjects, 6 were heterozygous and 1 was homozygous for the mutation. A hyperinsulinemic euglycemic clamp study was performed in 3 of 6 diabetic patients with the heterozygous Gly972Arg mutation and in 60 without it. Both groups showed almost the same levels of insulin sensitivity (glucose infusion rate, GIR = 50.2 +/- 3.0 vs. 51.3 +/- 12.1 micromol/kg/min). Similarly, there was no difference in insulin sensitivity between healthy subjects with and without the mutation using the homeostasis model assessment (HOMA index = 1.14 +/- 0.50 vs. 1.02 +/- 0.63). The frequency of the Gly972Arg allele was not increased in diabetic patients compared with control subjects even in aged (>50 years old) or obese (BMI >/=25) subjects. Among healthy subjects, we identified a 25-year-old male with the homozygous Gly972Arg allele. He was slightly obese (BMI = 25.5) but showed relatively high insulin sensitivity, almost equal to that of healthy subjects without the mutation (GIR = 67.2 vs. 71.8 +/- 22.0 micromol/kg/min). Because the GIR in healthy subjects was significantly higher compared with that in type-2 DM patients, we speculate that another genetic or environmental factor producing a more deleterious effect on insulin sensitivity may exist in diabetic patients. We conclude that this gene abnormality does not play a role in the pathogenesis of insulin resistance and type-2 DM.

Troglitazone reduces plasma levels of tumour necrosis factor-alpha in obese patients with type 2 diabetes

We evaluated the effect of troglitazone (given orally 400 mg/day) on glucose intolerance and on the plasma levels of tumour necrosis factor-alpha (TNF-alpha) in 12 obese patients with type 2 diabetes for 12 weeks. Troglitazone significantly decreased fasting plasma glucose, serum C-peptide, serum insulin and HbA1c levels. Plasma levels of TNF-alpha were significantly reduced by troglitazone administered for 8 and 12 weeks. Troglitazone administration significantly improved insulin resistance, but did not affect pancreatic beta-cell function as evaluated by the homeostasis model assessment (HOMA). In the present study, we reported for the first time that troglitazone administration significantly reduces plasma levels of TNF-alpha in obese patients with type 2 diabetes.

Plasma brain natriuretic peptide levels in normotensive noninsulin-dependent diabetic patients with microalbuminuria

Brain natriuretic peptide (BNP), a member of the natriuretic peptide family, is produced and released from cardiac ventricles. BNP regulates the body fluid volume, blood pressure, and vascular tones through the A-type guanylate cyclase-coupled receptor. The presence of renal dysfunction in patients with diabetes affects the plasma levels of atrial natriuretic peptide (ANP). In the present study, we investigated the plasma levels of BNP and ANP and their relationship in normotensive diabetic patients with normoalbuminuria and microalbuminuria. Forty-seven normotensive lean noninsulin-dependent diabetic patients (31 with normoalbuminuria, 16 with microalbuminuria), with normal cardiac function, and 30 age-matched control subjects were enrolled in this study. The plasma levels of BNP in diabetic patients with microalbuminuria were significantly higher than those in diabetic patients with normoalbuminuria (16.7+/-2.4 vs. 9.6+/-1.3 pg/mL, P<0.01) or normal subjects (16.7+/-2.4 vs. 7.0+/-0.6 pg/mL, P<0.01). There was a significant positive correlation between plasma BNP levels and urinary albumin excretion rate in all diabetic patients (r = 0.58, P<0.0001). There was also a significantly positive correlation between plasma BNP and ANP levels in diabetic patients (r = 0.62, P<0.0001). The increased plasma level of BNP in patients with microalbuminuria and its significant correlation with urinary albumin excretion rate suggest that the elevated circulating levels of BNP are caused by the presence of diabetic nephropathy. Down-regulation of A-type guanylate cyclase-coupled receptor of renal tubules may explain the increased plasma levels of both BNP and ANP in normotensive diabetic patients with microalbuminuria.

Codon 54 polymorphism of the fatty acid binding protein gene and insulin resistance in the Japanese population

AIM

To determine the relationship of the polymorphism at codon 54 of the intestinal fatty acid binding protein gene (FABP2) with insulin resistance and susceptibility to Type 2 diabetes mellitus (DM) in the Japanese population.

METHODS

We evaluated the polymorphism by the polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) in 150 Type 2 DM patients and 147 healthy control subjects. The frequency of alleles encoding threonine (Thr54) and alanine (Ala54) at codon 54 of FABP2 in Type 2 DM patients was compared with that of healthy controls. Insulin sensitivity was assessed by the hyperinsulinaemic euglycaemic clamp in Type 2 DM patients with Ala54 homozygotes, Ala54/Thr54 heterozygotes and Thr54 homozygotes and by homeostasis model assessment (HOMA) in the nondiabetic group.

RESULTS

The frequency of alleles encoding Ala54 and Thr54 was 0.59 and 0.41 in Type 2 DM patients, respectively, similar to that observed in nondiabetic controls (0.64 for Ala54 and 0.36 for Thr54). Insulin sensitivity was not significantly different between subjects with and without Thr54 allele either within the DM group or healthy controls.

CONCLUSIONS

The allele encoding threonine in the FABP2 does not predispose to Type 2 DM or insulin resistance in the Japanese population.

Significant decreased insulin secretion in a diabetic patient with clinically probable multiple sclerosis

A 39-year-old man with chief complaints of aphasia, disorientation and acalculia was admitted to our hospital. He was diagnosed as a clinically probable case of multiple sclerosis (MS) and his symptoms improved while on steroid pulse therapy. The patient had been diagnosed as having diabetes mellitus 16 years before the onset of MS and his insulin secretion further decreased with time. Slight insulin resistance was observed during a euglycemic hyperinsulinemic clamp study. These results suggested that this patient developed diabetes mellitus mainly due to the decrease of insulin secretion.

Serum levels of tumor necrosis factor-alpha are increased in obese patients with noninsulin-dependent diabetes mellitus

To clarify the significance of the serum levels of tumor necrosis factor-alpha (TNF-alpha) in the mechanism of insulin resistance, we studied 12 obese patients with noninsulin-dependent diabetes mellitus (NIDDM). We evaluated the relationship of TNF-alpha levels with the visceral, subcutaneous, and total fat areas measured by computed tomography (CT), and with insulin resistance evaluated by the glucose infusion rate (GIR) observed during an euglycemic hyperinsulinemic clamp study. Controls consisted of 12 normal subjects and 12 nonobese patients with NIDDM. TNF-alpha levels were measured using a high sensitivity enzyme-linked immunosorbent assay. Following admission, all patients with NIDDM participated in a 4-week program of diet and exercise. After this treatment, we evaluated the relationship of the serum levels of TNF-alpha with the area of body fat, the GIR, and the resultant change in the TNF-alpha level. Serum levels of TNF-alpha in the obese patients with NIDDM significantly exceeded those observed in normal subjects (P < 0.01) or in the nonobese patients with NIDDM (P < 0.01). Serum levels of TNF-alpha in obese NIDDM patients showed a significant positive correlation with the area of visceral fat before (r = 0.662, P < 0.03) and after (r = 0.508, P < 0.05) the treatment; similar correlation was observed in all patients with NIDDM before (r = 0.537, P < 0.02) and after (r = 0.430, P < 0.05) the treatment. Serum levels of TNF-alpha in obese NIDDM patients showed a significant negative correlation with GIR after the treatment (r = -0.595, P < 0.05). Serum levels of TNF-alpha were significantly reduced in the obese patients with NIDDM after the treatment (P < 0.01), while those in the nonobese NIDDM patients were unchanged. These results suggest that serum TNF-alpha levels may play an important role in mechanism of insulin resistance associated with obesity.

Acute and chronic regulation of serum sex hormone-binding globulin levels by plasma insulin concentrations in male noninsulin-dependent diabetes mellitus patients

This study was undertaken to investigate the relationship of the serum sex hormone-binding globulin (SHBG) levels with the plasma insulin concentration and with the insulin resistance in male subjects with noninsulin-dependent diabetes mellitus (NIDDM). This investigation comprised 12 patients with NIDDM and 16 normal subjects matched for age, sex, and body mass index (BMI). There was a significant increase in insulin levels (P < 0.03) and a decrease in SHBG levels (P < 0.01) in the diabetic group as compared with those of the normal group. The sex hormone and plasma insulin levels were measured in NIDDM patients undergoing exercise and dietary therapy. Insulin sensitivity was evaluated by the hyperinsulinemic euglycemic clamp technique expressed as the glucose infusion rate (GIR) before and after the treatment. The SHBG levels correlated significantly with the insulin concentrations (r = -0.643, P < 0.05) and with the GIR (r = 0.615, P < 0.05) before the treatment. The SHBG levels (P < 0.02) and GIR (P < 0.01) increased, and the insulin concentrations (P < 0.01) decreased significantly during the treatment. The SHBG levels showed a negative and significant correlation with the plasma insulin concentrations at the end of the clamp study before (r = -0.615, P < 0.05) and after (r = -0.626, P < 0.05) the treatment. These findings suggest that, in the hyperinsulinemic state, plasma insulin has a direct effect on the SHBG levels. SHBG levels decreased significantly during the clamp study before (P < 0.02) and after (P < 0.01) the treatment. This may represent the acute effect of insulin on the SHBG levels. Briefly, these results suggest that insulin may directly affect the SHBG levels and that SHBG may constitute an index of the insulin resistance only in the hyperinsulinemic state.