Body Mass Index, Fasting Plasma Glucose Levels, and C-peptide Levels as Predictors of the Future Insulin Use in Japanese Type 2 Diabetic Patients

Transcription

1 Endocrine Journal 2010, 57 (3), Or i g i n a l Body Mass Index, Fasting Plasma Glucose Levels, and C-peptide Levels as Predictors of the Future Insulin Use in Japanese Type 2 Diabetic Patients Atsushi Goto 1), 3), Maki Takaichi 1), Miyako Kishimoto 1), Yoshihiko Takahashi 1), Hiroshi Kajio 1), Takuro Shimbo 2) and Mitsuhiko Noda 1) 1) Department of Diabetes and Metabolic Medicine, International Medical Center of Japan, Toyama Hospital, Tokyo, Japan 2) Department of Clinical Research and Informatics, Research Institute, International Medical Center of Japan, Tokyo, Japan 3) Department of Epidemiology, School of Public Health, University of California, Los Angeles, United States Abstract. Insulin therapy represents the most effective and reliable means of achieving satisfactory glycemic control. However, few studies have evaluated the predictors of future insulin use. The purpose of this study was to investigate the predictors of future insulin use in type 2 diabetic patients. In this study, we conducted a chart review of 158 Japanese type 2 diabetic patients admitted to our hospital for stringent glycemic control. Of the 158 subjects, 92 satisfied the inclusion criteria for this study. We assessed the associations between baseline BMI, fasting plasma glucose levels (FPG) and serum and urinary C-peptide levels (scpr and ucpr), and insulin usage at 6 months after discharge. We also computed the area under the curve (AUC) in receiver operating characteristic (ROC) curve for each predictor to predict the future insulin use. After adjustment for gender, age, and BMI, the multivariable odds ratios (ORs) for future insulin use in the highest tertile as compared with lowest tertile were 0.12 for BMI (95% confidence interval [CI], ), 17.0 for FPG (95% CI, ), 0.12 for scpr (95% CI, ), and 0.03 for ucpr (95% CI, ). Prediction analyses showed that the AUCs for BMI, FPG, scpr, and ucpr were 0.73, 0.76, 0.74, and 0.78, respectively, which suggests that the predictive abilities of these predictors do not differ substantially. In conclusion, this study suggests that BMI, FPG, scpr, and ucpr are strong predictors of the future insulin use in type 2 diabetic patients. Key words: Type 2 diabetes mellitus, Insulin therapy, Body mass index, Fasting plasma glucose levels, C-peptide levels Although the increasing availability of new classes of blood glucose-lowering medications has increased the treatment options available for type 2 diabetic patients, insulin therapy still serves as the most effective and reliable means of achieving glycemic goals. In practice, the decision to start a type 2 diabetic patient on insulin therapy is based on the glycemic control status, as assessed by the fasting plasma glucose levels (FPG) and HbA 1c [1, 2]. As it is widely known that both impaired insulin secretion and in- Received Sep. 24, 2009; Accepted Dec. 3, 2009 as K09E-279 Released online in J-STAGE as advance publication Dec. 23, 2009 Correspondence to: Mitsuhiko Noda, M.D., Ph.D., Director, Department of Diabetes and Metabolic Medicine, International Medical Center of Japan, Toyama , Shinjuku-ku, Tokyo , Japan. mnoda@imcj.hosp.go.jp Conflicts of Interest: There are no conflicts of interests to declare. sulin resistance play critical roles in the development of hyperglycemia in patients of type 2 diabetes, measurement of the serum C-peptide levels (scpr) and urinary C-peptide levels (ucpr), which are known to be useful markers of beta-cell function, may also provide important information for the choice of therapy [3-13]. However, few studies have been conducted to identify factors that might determine or predict the therapy of choice in type 2 diabetic patients. To identify the predictors of the future insulin use, we comprehensively examined the association between various clinical characteristics measured during the course of hospitalization, and usage of insulin at 6 months after discharge from the hospital in patients with type 2 diabetes. We also conducted prediction analyses using receiver operating characteristic curves to assess the predictive performance of each predictor.

2 238 GOTO et al. Methods Study subjects In this study, we reviewed the medical records of all patients who were admitted to the department of diabetes and metabolic medicine at the International Medical Center of Japan Hospital, a 700-bedded teaching hospital, between April 2006 and March Subjects were selected from among type 2 diabetic patients aged 18 years or over who were admitted to our hospital for stringent control of blood glucose. A total of 158 patients were admitted from April 2006 to March Of the 158 patients, we excluded 66 patients; 30 were already receiving insulin therapy at the time of admission, 11 were lost to follow-up at our hospital, 9 had evidence of renal dysfunction [serum creatinine 1.5 mg/dl], 9 had malignancy, 4 had liver cirrhosis, and 3 had an acute infection. We included the remaining 92 (58.6%) patients in this study. Laboratory evaluations From each study participant, blood samples were collected at admission after the subject had fasted for at least 8 hours. Also, 24-hour urine samples were collected at admission and stored in a refrigerator. Fasting plasma glucose (FPG) concentrations were measured by the glucose electrode method. HbA 1 c levels were measured by high-pressure liquid chromatography (HPLC). The plasma total cholesterol and triglyceride levels were measured by enzymatic assays, and the serum HDL cholesterol levels were measured by the direct method. Serum and urinary C-peptide levels were determined by the electrochemiluminescence immunoassay (ECLIA) method. Assessment of clinical characteristics At the time of admission, a careful history was obtained from the patients about their use of oral hypoglycemic agents, the time of diagnosis of type 2 diabetes mellitus, family history of type 2 diabetes mellitus, and the smoking history. We measured the heights, weights, and systolic and diastolic blood pressure levels (mmhg) at admission. Patients were followed up clinically and the usage of insulin at 6 months after discharge was ascertained from the medical records. Statistical analysis Baseline characteristics of the participants were compared according to the usage of insulin at 6 months after discharge using chi-square tests or fisher s exact tests for discrete variables, and t-tests of equal group means for continuous variables. To assess the association between the clinical characteristics and future insulin use, we selected body mass index (BMI), fasting plasma glucose (FPG), serum C-peptide levels (scpr), CPR index (C-peptide immunoreactivity index, 100 x scpr divided by FPG), and urinary C-peptide levels (ucpr) as predictors based on prior biological knowledge and univariate associations[11]. We categorized BMI, FPG, scpr, CPR index, and ucpr into tertiles. Then, we performed multiple logistic regression analyses to estimate the odds ratios (ORs) and 95% confidence intervals (CI) for future insulin use for each tertile of each predictor using the lowest tertile as the reference category. We tested for a linear trend across increasing tertiles of the predictor variables by assigning a median value for each tertile treated as a continuous variable. Though categorical tertile analyses avoid making the assumption of linear relation, such grouping discards information and is likely to lose statistical power. Thus, we also conducted logistic regression analyses treating the predictors as continuous variables. Then, we estimated the ORs per 1 SD increase in each of predictors, assuming a linear relationship. Finally, we conducted prediction analyses using receiver operating characteristic curves to assess the predictive performance of single predictors (BMI, FPG, scpr, CPR index, and ucpr). We computed the area under the curve (AUC), the number of true-positive, false-positive, true-negative and false-negative, sensitivity, specificity, positive predictive value, and negative predictive value of each predictor using various cutoff points to predict future insulin use. All P values were two-tailed and all statistical analyses were conducted using STATA (version 10.1; StataCorp, Texas). Results The baseline characteristics of the study participants according to the usage of insulin at 6 months after discharge are shown in Table 1. Patients with insulin treatment at 6 months after discharge from the hospital tended to include a greater number of males and subjects receiving pioglitazone, insulin therapy during hospitalization, and insulin therapy at discharge, a smaller number of subjects receiving bigu-

3 Predictors of the future insulin use 239 Table 1. Baseline characteristics according to the use of insulin at 6 months after discharge Insulin (+) Insulin (-) P value b N Male (%) 76.9 (20/26) 57.6 (28/66) Age (years) 62.1 ± ± Years since Diagnosis of DM (years) 9.54 ± ± Family History of DM (%) 53.8 (14/26) 62.1 (41/66) Smoking History (%) 61.5 (16/26) 57.6 (28/66) SU (%) 61.5 (16/26) 65.2 (43/66) Biguanides (%) 19.2 (5/26) 34.8 (23/66) Pioglitazone (%) 34.6 (9/26) 19.7 (13/66) Insulin use during hospitalization (%) 100 (26/26) 22.7 (15/66) <0.001 Insulin use at discharge (%) 96.3 (25/26) 3.0 (2/66) <0.001 BMI (kg/m 2 ) 22.3 ± ± Systolic BP (mmhg) ± ± Diastolic BP (mmhg) 76.5 ± ± FPG (mg/dl) ± ± HbA 1 c (%) 10.9 ± ± T-Chol (mg/dl) ± ± HDL-C (mg/dl) 70.0 ± ± TG (mg/dl) ± ± Serum C-peptide (ng/ml) 1.4 ± ± Urinary C-peptide (µg/day) 53.7 ± ± 62.8 <0.001 Abbreviations: DM, type 2 diabetes mellitus; SU, sulfonylurea agent; BMI, body mass index; BP, blood pressure; FPG, fasting plasma glucose levels; TG, serum triglyceride levels. a Unless otherwise indicated, values are expressed as the mean (± SD). b A chi-square test or fisher s exact test was used for discrete variables, and a t-test of equal group means was used for continuous variables. independent variables (Model 2), the odds ratios (ORs) for insulin use in the highest tertile compared with the lowest tertile were 0.12 for BMI (95% CI, ; P for trend, 0.002), 17.0 for FPG (95% CI, ; P for trend, 0.001), 0.12 for scpr (95% CI, ; P for trend, 0.018), 0.09 for CPR index (95% CI, ; P for trend, 0.008), and 0.03 for ucpr (95% CI, ; P for trend, <0.001). Additional adjustment for FPG attenuated, but did not entirely eliminate these associations (Model 3). BMI and FPG were significantly associated with insulin use even after further adjustment for ucpr (Model 4). Further analyses, treating the predictors as continuous variables, showed similar results. We further conducted prediction analyses using reanides, higher values of FPG, HbA 1 c, and HDL-C, and lower values of BMI, scpr, and ucpr at the baseline than patients without insulin treatment at 6 months. We did not observe an association between years since diagnosis of type 2 diabetes and future insulin use. To further examine the associations between the predictors and the future use of insulin, we categorized BMI, FPG, scpr, CPR index, and ucpr into tertiles, and performed multiple logistic regression analyses to estimate the odds ratios (ORs) and 95% CI for receiving insulin treatment at 6 months (Table 2). In crude model (Model 1), all predictors were associated with the future insulin use. After adjustment for gender (male or female), age (continuous) and BMI (continuous), which were not examined as the primary

4 240 GOTO et al. Table 2. Fasting plasma glucose and C-peptide levels, and the risk of future insulin use among type 2 diabetic patients Tertile1 (lowest) Tertile 2 Tertile 3 (highest) P for Continuous trend (per 1 SD increase) ** BMI Median (Range), kg/m ( ) 24.8 ( ) 28.8 ( ) N (Insulin (+) / Insulin (-)) 8/5 11/25 7/36 Odds Ratio: Model 1 a ( ) 0.12 ( ) ( ) Odds Ratio: Model 2 b ( ) 0.12 ( ) ( ) Odds Ratio: Model 3 c ( ) 0.13 ( ) ( ) Odds Ratio: Model 4 d ( ) 0.30 ( ) ( ) FPG Median (Range), mg/dl 130 (80-161) 197 ( ) 269 ( ) N (Insulin (+) / Insulin (-)) 3/28 9/24 14/14 Odds Ratio: Model 1 a ( ) 9.33 ( ) ( ) Odds Ratio: Model 2 b ( ) 17.0 ( ) ( ) Odds Ratio: Model 4 d ( ) 7.41 ( ) ( ) scpr Median (Range), ng/dl 1.1 ( ) 1.8 ( ) 2.8 ( ) N (Insulin (+) / Insulin (-)) 11/15 7/18 2/23 Odds Ratio: Model 1 a ( ) 0.12 ( ) ( ) Odds Ratio: Model 2 b ( ) 0.12 ( ) ( ) Odds Ratio: Model 3 c ( ) 0.21 ( ) ( ) CPR index Median (Range) 0.4 ( ) 1.0 ( ) 1.7 ( ) N (Insulin (+) / Insulin (-)) 12/14 6/19 2/23 Odds Ratio: Model 1 a ( ) 0.10 ( ) ( ) Odds Ratio: Model 2 b ( ) 0.09 ( ) ( ) Odds Ratio: Model 3 c ( ) 0.19 ( ) ( ) ucpr Median (Range), µg/day 38.4 ( ) 78.9 ( ) ( ) N (Insulin (+) / Insulin (-)) 16/15 9/21 1/29 Odds Ratio: Model 1 a ( ) 0.03 ( ) < ( ) Odds Ratio: Model 2 b ( ) 0.03 ( ) < ( ) Odds Ratio: Model 3 c ( ) 0.04 ( ) ( ) * Abbreviations: BMI, body mass index; FPG, fasting plasma glucose levels; scpr, serum C-peptide levels; CPR index, C-peptide immunoreactivity index, 100 x scpr divided by FPG. ** ORs based on per 1 SD increase in BMI, FPG, scpr, CPR index, and ucpr as continuous variables (each SD was 5.81 kg/m 2 for BMI, 68.5 mg/dl for FPG, 1.3 ng/dl for scpr, 0.88 for CPR index, and 60.4 µg/day for ucpr.) a Model 1: crude model. b Model 2 adjusted for gender (male or female), age (continuous) and BMI (continuous) that were not examined as the primary independent variable. c Model 3 adjusted for variables in Model 2 plus the fasting plasma glucose levels (continuous). d Model 4 adjusted for variables in Model 3 plus the urinary C-peptide levels (continuous). ceiver operating characteristic curves to assess the predictive performance of single predictors (Table 3). Although the CPR index and ucpr showed the highest AUCs, the AUCs of these predictors were not markedly different from one another. CPR index showed higher AUCs than scpr, which suggests that the predictive performance of scpr could be improved by combining it with FPG. Diagnostic measures, including the sensitivity, specificity, positive predictive values and negative predictive values were substantially dependent on the chosen cutoff points.

5 Predictors of the future insulin use 241 Table 3. AUCs, the number of true-positive, false-positive, true-negative, and false-negative, and sensitivity, specificity, positive predictive value, and negative predictive value using various cutoff points for the prediction of future insulin use Predictors, Cutoff point BMI (kg/m 2 ) 0.73 AUC TP / FP TN / FN Sensitivity (%) Specificity (%) PPV (%) NPV (%) /5 61/ /14 52/ /30 36/ /46 20/ /57 9/ FPG (mg/dl) /47 19/ /32 34/ /22 44/ /7 59/ /2 64/ scpr (ng/ml) /5 51/ /15 41/ /24 32/ /30 26/ /40 16/ CPR index /8 48/ /16 40/ /22 44/ /31 25/ /34 22/ ucpr (µg/day) /10 55/ /17 48/ /27 38/ /32 33/ /31 34/ Abbreviations: BMI, body mass index; FPG, fasting plasma glucose levels; scpr, serum C-peptide levels; CPR index, C-peptide immunoreactivity index, 100 x scpr divided by FPG; AUC, area under the curve; TP, the number of true positive; FP, the number of false positive; TN, the number of true negative; FN, the number of false negative; PPV, positive predictive value; NPV, negative predictive value. Discussion There are two main findings of the present study conducted in type 2 diabetic patients. First, we found that lower values of the body mass index, elevated fasting plasma glucose levels, decreased serum and urinary C-peptide levels, and decreased CPR index were associated with future insulin use. These associations were independent of possible confounding factors, including age and gender. Secondly, the present study showed the predictive abilities of body mass index, fasting plasma glucose levels, serum and urinary

6 242 GOTO et al. C-peptide levels, and the CPR index for the future insulin use. Although previous studies have investigated the association between markers of beta-cell function and future insulin use, to the best of our knowledge, none has comprehensively examined the relationships of various clinical variables, including body mass index, fasting plasma glucose levels, and serum and urinary C-peptide levels with future insulin use [4, 5, 7, 8, 11, 14-16]. As expected, we found that increased FPG was associated with future insulin use. This observed association was significant even after adjustment for clinical characteristics and the ucpr. As a large body of evidence has shown that elevated plasma glucose concentrations are positively associated with adverse outcomes, and that intensive plasma glucose control significantly reduces the incidence of microvascular complications in type 2 diabetic patients, current guidelines for the treatment of type 2 diabetes recommend initiation of insulin therapy if target glycemic goals are not achieved with other treatment measures, including lifestyle interventions and oral hypoglycemic therapy [1, 2, 17-19]. Therefore, the observed positive association between FPG and insulin use is not surprising. Our findings indicated that lower values of BMI were associated with future insulin use. Nonetheless, it is important to emphasize that this finding does not imply that increased adiposity itself is protective against the future use of insulin therapy. This association can be explained by two possible mechanisms. Firstly, lean patients usually have less margin for reduction of adiposity to further decrease relative insulin resistance. Secondly, lower BMI could be a surrogate marker for impaired insulin secretion which is in most part genetically programmed [16, 20]. Thus, patients with lower BMI might be more likely to need insulin treatment to achieve target blood glucose levels. Our finding of the associations between the scpr, ucpr, and CPR index and future insulin use are consistent with previous reports. Asano et al. reported that the serum and urinary C-peptide levels corrected for the fasting plasma glucose were associated with the types of antidiabetic therapy that were used at the end of the observation period, and concluded that these indices were useful for decision-making as to insulin therapy in patients with type 2 diabetes [11]. The clinical practice of measuring the serum and urinary C-peptide levels to help guide the choice of ther- apy in patients with type 2 diabetes is based on the known pathophysiology of type 2 diabetes. In theory, the lower the C-peptide levels, the more appropriate the use of insulin or insulin secretagogues might be as the treatment of choice. To investigate this hypothesis, Ko et al. followed 503 patients with type 2 diabetes and reported that phenotype-targeted therapy based on the fasting C-peptide levels might improve the clinical outcomes in type 2 diabetic patients [13]. We usually make the decision of whether insulin therapy should be initiated based on the glycemic control status [1, 2]. Thus, it is essential to assess whether the predictive abilities of BMI and C-peptide levels might be superior to that of the glycemic control status. The AUCs for single predictors are shown in Table 3. The AUCs for FPG, BMI, scpr, CPR index, and ucpr were 0.76, 0.73, 0.74, 0.78, and 0.78, respectively. These findings indicate that BMI, and C-peptide levels may not substantially improve upon the predictive ability of the glycemic control status for future insulin use. In order to apply the beneficial value of predictors in decision making in clinical settings, we need to choose a cutoff point for each predictor. As shown in Table 3, diagnostic measures, including the sensitivity, specificity, positive predictive values, and negative predictive values depend on the chosen cutoff points. Importantly, whether we want a higher sensitivity or higher specificity depends entirely on the clinical setting. Therefore, we have not defined cutoff points for predictors. Instead, we encourage clinicians to choose cutoff points depending on the clinical setting. Several issues with regard to our study design need to be acknowledged. First of all, the sample size was relatively small. This is because we implemented strict exclusion criteria. We excluded 30 patients who were already using insulin at the time of admission, whereby we presumably reduced potential biases due to the inclusion of patients using insulin at baseline. Second, there were 11 patients who were lost to follow-up, which might have introduced a selection bias. Third, our findings could have been affected by unmeasured confounding factors. Nonetheless, it is unlikely that the observed associations in this study would be eliminated by controlling for unmeasured confounding factors, because the magnitudes of the associations were strong. Fourth, it is possible that physician s choice of insulin therapy might have been affected by the beta cell function indices. However,

7 Predictors of the future insulin use 243 the decision criteria for insulin use of our hospital is based on the current treatment guide for type 2 diabetes, in which initiation of insulin therapy is recommended if target glycemic goals are not achieved with other treatment measures, including lifestyle interventions and oral hypoglycemic therapy [1, 2, 17-19]. Finally, because the decision to start insulin therapy might differ across populations and because the degree of adiposity in our study population is different from other ethnic groups, it might not be possible to gener- alize these results to other populations. In conclusion, lower values of the body mass index, elevated fasting plasma glucose, and decreased serum and urinary C-peptide levels were associated with future insulin use in our Japanese type 2 diabetic patients. These results might help clinicians to tell their type 2 diabetic patients about the predicted clinical course, and choose the appropriate therapy. Further studies are warranted to confirm these findings. References 1. Japan Diabetes Society: Treatment Guide for Diabetes Nathan DM, Buse JB, Davidson MB, Ferrannini E, Holman RR, Sherwin R, Zinman B (2009) Medical management of hyperglycemia in type 2 diabetes: a consensus algorithm for the initiation and adjustment of therapy: a consensus statement of the American Diabetes Association and the European Association for the Study of Diabetes. Diabetes Care 32: Peacock I, Tattersall RB (1984) The difficult choice of treatment for poorly controlled maturity onset diabetes: tablets or insulin? Br Med J (Clin Res Ed) 288: Aro A (1985) Role of plasma C-peptide determination in the management of type II diabetes. Acta Endocrinol Suppl (Copenh) 272: Koskinen P, Viikari J, Irjala K, Kaihola HL, Seppala P (1985) C-peptide determination in the choice of treatment in diabetes mellitus. Scand J Clin Lab Invest 45: Faber OK, Binder C (1986) C-peptide: an index of insulin secretion. Diabetes Metab Rev 2: Kyllastinen M, Elfving S (1986) Serum C-peptide concentrations and their value in evaluating the usefulness of insulin therapy in elderly diabetics. Gerontology 32: Nielsen NV, Tronier B (1986) C-peptide in diabetes mellitus treated with insulin. A 3-year epidemiological study on the island of Falster, Denmark. Diabetes Res 3: Keller U, Pasquel M, Berger W (1987) [C-peptide determination in diabetics for the evaluation of insulin requirements]. Schweiz Med Wochenschr 117: Sacks DB, Bruns DE, Goldstein DE, Maclaren NK, McDonald JM, Parrott M (2002) Guidelines and recommendations for laboratory analysis in the diagnosis and management of diabetes mellitus. Clin Chem 48: Asano T, Kawamura M, Watanabe T, Abe M, Chin R, Miyazaki S, Hirata Y (2008) Indices of Urinary and Serum C-peptide Corrected with Fasting Plasma Glucose for Decision-making of Insulin Therapy in Type 2 Diabetes-Validation and Comparison. J. Japan Diab. Soc. 51: Muoio DM, Newgard CB (2008) Mechanisms of disease: molecular and metabolic mechanisms of insulin resistance and beta-cell failure in type 2 diabetes. Nat Rev Mol Cell Biol 9: Ko GT, So WY, Tong PC, Chan WB, Yang X, Ma RC, Kong AP, Ozaki R, Yeung CY, Chow CC, Chan JC (2009) Effect of interactions between C peptide levels and insulin treatment on clinical outcomes among patients with type 2 diabetes mellitus. CMAJ 180: Nielsen NV, Tronier B (1987) C-peptide and insulin secretion in diabetes mellitus treated with oral hypoglycaemic agents or diet alone. A 3 years epidemiological cohort study on the Island of Falster, Denmark. Diabetes Res 4: Chan WB, Chan JC, Chow CC, Yeung VT, So WY, Li JK, Ko GT, Ma RC, Cockram CS (2000) Glycaemic control in type 2 diabetes: the impact of body weight, beta-cell function and patient education. QJM 93: Chan WB, Tong PC, Chow CC, So WY, Ng MC, Ma RC, Osaki R, Cockram CS, Chan JC (2004) The associations of body mass index, C-peptide and metabolic status in Chinese Type 2 diabetic patients. Diabet Med 21: Ohkubo Y, Kishikawa H, Araki E, Miyata T, Isami S, Motoyoshi S, Kojima Y, Furuyoshi N, Shichiri M (1995) Intensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study. Diabetes Res Clin Pract 28: (1998) Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treat-

8 244 GOTO et al. ment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet 352: Meigs JB, Mittleman MA, Nathan DM, Tofler GH, Singer DE, Murphy-Sheehy PM, Lipinska I, D Agostino RB, Wilson PW (2000) Hyperinsulinemia, hyperglycemia, and impaired hemostasis: the Framingham Offspring Study. JAMA 283: Relimpio F, Losada F, Pumar A, Garcia de Pesquera F, Morales F, Acosta D, Astorga R (1997) Relationships of C-peptide levels and the C-peptide/bloodsugar ratio with clinical/biochemical variables associated with insulin resistance in orally-treated, well-controlled type 2 diabetic patients. Diabetes Res Clin Pract 36: