Breakout: Initiation and Titration of Insulin

1. Breakout: Initiation and Titration of Insulin Martin J. Abrahamson, MD Medical Director and Senior Vice President Joslin Diabetes Center Associate Professor of Medicine Harvard Medical School Boston, Massachusetts Martha M. Funnell, MS, RN, CDE Michigan Diabetes Research and Training Center Juvenile Diabetes Research Foundation Center for the Study of the Complications in Diabetes Ann Arbor, Michigan

2. Learning Objectives Understand the natural history of type 2 diabetes Understand the normal physiology of insulin secretion and recognize the different insulins used to treat type 2 diabetes Develop an algorithm for the initiation and intensification of insulin treatment in type 2 diabetes

3. Case Study: I.S. I have this 66 year old male who has had type 2 diabetes for about 6 years. He is taking 2 oral medications for his diabetes and his control has deteriorated again. Do you think he is just not watching his diet? Is it normal for someone’s glucose control to deteriorate like this? Has he developed type 1 diabetes now? Should I start him on insulin? If so what is the best insulin to use, in other words how should I do it?

4. Case Study: I.S. 66 year old male Type 2 diabetes for 6 years Co morbid conditions Hypertension Hyperlipidemia Meds Glimepiride 4 mg daily Metformin 1 g bid Simvastatin 20 mg qd Lisinopril 40 mg qd HCTZ 12.5 mg qd Amlodipine 5 mg qd

5. Case Study: I.S. Non proliferative retinopathy No known neuropathy or nephropathy No known CAD or PVD A1C 8.8% A1C 7% a year previously A1C 6.5% soon after 2 agents started (2 years ago) Consistent carbohydrate intake Walks 4 times per week for 30 mins each time

6. Case Study: I.S. Physical examination Height 68” Weight 203 lbs BMI 30 kg/m2 Acanthosis nigricans BP 120/70 Pulse 66 Background retinopathy but no other evidence of any micro or macrovascular complications

7. Case Study: I.S. Laboratory Data Fasting glucose 165 mg/dL A1C 8.8% Lipids TC 175 mg/dL Triglycerides 142 mg/dL HDL 44 mg/dL LDL 103 mg/dL Urine albumin: creatinine 2.1 mcg/mg LFTs normal Creatinine 0.9 mg/dL

8. Physiologic Insulin Secretion: 24-hour Profile When we consider glucose control, we need to focus on both postprandial and basal requirements. This slide illustrates the normal physiologic response of glucose and insulin to meals, which highlights the need for both basal and meal-time insulin. Meal insulin release occurs in response to nutrient ingestion. Basal insulin is continuously secreted over a 24-hour period. In the past, we have had to make due with various insulin formulations that did not have adequate pharmacokinetics to duplicate these profiles. However, within the past few years, new insulin analogs that provide more physiologic profiles have been developed. When we consider glucose control, we need to focus on both postprandial and basal requirements. This slide illustrates the normal physiologic response of glucose and insulin to meals, which highlights the need for both basal and meal-time insulin. Meal insulin release occurs in response to nutrient ingestion. Basal insulin is continuously secreted over a 24-hour period. In the past, we have had to make due with various insulin formulations that did not have adequate pharmacokinetics to duplicate these profiles. However, within the past few years, new insulin analogs that provide more physiologic profiles have been developed.

9. Insulins Available for Clinical Use Prandial insulin Short-acting (human): regular Rapid-acting (analog): aspart, lispro, glulisine Basal insulin Intermediate-acting: NPH Long-acting analogs: detemir, glargine Premixed insulins Provide both basal and mealtime coverage Human: Humulin 70/30 Analogs: Humalog 75/25; Humalog 50/50 Novolog 70/30

11. Self-Monitoring of Blood Glucose T1DM and pregnant women taking insulin- = 4X/day T2DM on oral agent therapy- sufficient to facilitate reaching glucose goals T2DM taking insulin- more frequently than those not using insulin Patients using multiple insulin injections- = 4X/day For patients using less frequent insulin injections or oral agents or medical nutrition therapy alone, SMBG is useful in achieving glycemic goals

12. Self-Monitoring of Blood Glucose (cont.) When adding to or modifying therapy, T1DM and T2DM patients should test more often than usual The role of SMBG in stable diet–treated patients with T2DM is not known To achieve postprandial glucose targets, postprandial SMBG may be appropriate Instruct the patient in SMBG and routinely evaluate the patient’s technique and ability to use data to adjust therapy

13. Let’s Get More Data… Current diabetes medications: Glimepiride 4 mg daily Metformin 1 g bid Patient I.S. is testing sugars in the morning before breakfast only We ask him to test sugars twice daily, fasting and 2 hours after meals, on a “grid system” One week later he is seen in the office with the following glucose readings:

15. Let’s Get More Data… What are his thoughts/concerns about starting insulin? Common Fears: Needles Hypoglycemia Weight gain Common Beliefs: Insulin does not work Insulin causes complications Insulin is a personal failure Adverse impact on relationships/lifestyle

16. 24-hour Plasma Glucose Curve:Rationale for Adding Basal Insulin

17. 24-hour Plasma Glucose Curve:Rationale for Adding Basal Insulin

18. Starting Basal Insulin In Type 2 DM – What I do Continue oral agent(s) at same dose Add single bedtime insulin dose (10 U or 0.1 U/kg) NPH Insulin detemir Insulin glargine Adjust dose according to fasting glucose (FBG) Increase insulin dose every 3 to 5 days as needed ? 2 units if FBG >120 mg/dL ? 4 units if FBG >140 mg/dL ? 6 units if FBG >160 mg/dL

19. Treat-to-Target Study: Mean Fasting Plasma Glucose During Study

21. Insulin Glargine vs. NPH InsulinAdded to Oral Therapy Similar trends were observed when the overall number of symptomatic hypoglycemic episodes occurring over time were analyzed. Between midnight and 8 am, and breakfast time, there were fewer episodes of symptomatic hypoglycemia in the insulin glargine group compared with the NPH insulin group At other times of day, the number of episodes of symptomatic hypoglycemia were similar in both treatment groups ROSENSTOCK Abstract OBJECTIVE: To determine the safety and efficacy of the long-acting analog insulin glargine compared with NPH insulin in patients with type 2 diabetes who were previously treated with insulin alone. RESEARCH DESIGN AND METHODS: A total of 518 subjects with type 2 diabetes who were receiving NPH insulin with or without regular insulin for postprandial control were randomized to receive insulin glargine (HOE 901) once daily (n = 259) or NPH insulin once or twice daily in = 259) for 28 weeks in an open-label, multicenter trial. Doses were adjusted to obtain target fasting glucose <6.7 mmol/l. At study end point, the median total daily insulin dose in both treatment groups was 0.75 IU/kg. RESULTS: The treatment groups showed similar improvements in A1C from baseline to end point on intent-to-treat analysis. The mean change (means +/- SD) in A1C from baseline to end point was similar in the insulin glargine group (-0.41 +/- 0.1%) and the NPH group (-0.59 +/- 0.1%) after patients began with an average baseline A1C of approximately 8.5%. The treatments were associated with similar reductions in fasting glucose levels. Overall, mild symptomatic hypoglycemia was similar in insulin glargine subjects (61.4%) and NPH insulin subjects (66.%) However, nocturnal hypoglycemia in the insulin glargine group was reduced by 25% during the treatment period after the dose-titration phase(26.5 vs. 35.5%, P = 0.0136). Subjects in the insulin glargine group experienced less weight gain than those in the NPH group (0.4 vs. 1.4 kg, P < 0.0007). CONCLUSIONS: In patients with type 2 diabetes, once-daily bedtime insulin glargine is as effective as once- or twice-daily NPH in improving and maintaining glycemic control. In addition, insulin glargine demonstrates a lower risk of nocturnal hypoglycemia and less weight gain compared with NPH insulin. Similar trends were observed when the overall number of symptomatic hypoglycemic episodes occurring over time were analyzed. Between midnight and 8 am, and breakfast time, there were fewer episodes of symptomatic hypoglycemia in the insulin glargine group compared with the NPH insulin group At other times of day, the number of episodes of symptomatic hypoglycemia were similar in both treatment groups ROSENSTOCK Abstract OBJECTIVE: To determine the safety and efficacy of the long-acting analog insulin glargine compared with NPH insulin in patients with type 2 diabetes who were previously treated with insulin alone. RESEARCH DESIGN AND METHODS: A total of 518 subjects with type 2 diabetes who were receiving NPH insulin with or without regular insulin for postprandial control were randomized to receive insulin glargine (HOE 901) once daily (n = 259) or NPH insulin once or twice daily in = 259) for 28 weeks in an open-label, multicenter trial. Doses were adjusted to obtain target fasting glucose <6.7 mmol/l. At study end point, the median total daily insulin dose in both treatment groups was 0.75 IU/kg. RESULTS: The treatment groups showed similar improvements in A1C from baseline to end point on intent-to-treat analysis. The mean change (means +/- SD) in A1C from baseline to end point was similar in the insulin glargine group (-0.41 +/- 0.1%) and the NPH group (-0.59 +/- 0.1%) after patients began with an average baseline A1C of approximately 8.5%. The treatments were associated with similar reductions in fasting glucose levels. Overall, mild symptomatic hypoglycemia was similar in insulin glargine subjects (61.4%) and NPH insulin subjects (66.%) However, nocturnal hypoglycemia in the insulin glargine group was reduced by 25% during the treatment period after the dose-titration phase(26.5 vs. 35.5%, P = 0.0136). Subjects in the insulin glargine group experienced less weight gain than those in the NPH group (0.4 vs. 1.4 kg, P < 0.0007). CONCLUSIONS: In patients with type 2 diabetes, once-daily bedtime insulin glargine is as effective as once- or twice-daily NPH in improving and maintaining glycemic control. In addition, insulin glargine demonstrates a lower risk of nocturnal hypoglycemia and less weight gain compared with NPH insulin.

22. Treat-to-Target Study: Glargine vs. NPH Added to Oral Therapy for Patients with T2D Significant proportion of patients still not meeting A1C treatment goals Without addressing postprandial glucose, basal insulin is often not enough Treat-to-Target Study: Glargine vs. NPH Added to Oral Therapy for Patients With Type 2 Diabetes Riddle M, et al. Diabetes. 2002;Abstract 457. [CANNOT COMPLETE CITATION ON INFORMATION PROVIDED] Riddle MC, Rosenstock J, Gerich J; for the Insulin Glargine 4002 Study Investigators. The treat-to-target trial: randomized addition of glargine or human NPH insulin to oral therapy of type 2 diabetic patients. Diabetes Care. 2003;26:3080–3086. Treat-to-Target Study: Glargine vs. NPH Added to Oral Therapy for Patients With Type 2 Diabetes Riddle M, et al. Diabetes. 2002;Abstract 457. [CANNOT COMPLETE CITATION ON INFORMATION PROVIDED] Riddle MC, Rosenstock J, Gerich J; for the Insulin Glargine 4002 Study Investigators. The treat-to-target trial: randomized addition of glargine or human NPH insulin to oral therapy of type 2 diabetic patients. Diabetes Care. 2003;26:3080–3086.

23. Detemir vs. NPH in Type 2 Diabetes With a Treat-to-Target Design The pre-breakfast glycaemic target was reached by 63% vs. 68% (Idet vs. NPH) (NS) Reduction in HbA1C was 1.8% vs. 1.9% (Idet vs. NPH) The plot is using arithmetic means (results in previous slides show least square means corrected for baseline). Data V1 V2 V10 V13 Detemir 8.71 8.61 7.03 6.75 (6.58) NPH 8.66 8.51 6.82 6.62 (6.46) Already after 12 weeks of treatment patients are achieving a nice level of HbA1c with the treat to target. Getting it closer, +SD HbA1c levelThe pre-breakfast glycaemic target was reached by 63% vs. 68% (Idet vs. NPH) (NS) Reduction in HbA1C was 1.8% vs. 1.9% (Idet vs. NPH) The plot is using arithmetic means (results in previous slides show least square means corrected for baseline). Data V1 V2 V10 V13 Detemir 8.71 8.61 7.03 6.75 (6.58) NPH 8.66 8.51 6.82 6.62 (6.46) Already after 12 weeks of treatment patients are achieving a nice level of HbA1c with the treat to target. Getting it closer, +SD HbA1c level

24. Detemir Achieved Glycemic ControlWith a Low Rate of Hypoglycemia vs. Glargine The following is to be answered only in response to an unsolicited question. Levemir® demonstrated comparable A1c with less major and nocturnal hypoglycemia. The risk of major and nocturnal (11 pm to 6 am) hypoglycemia was 72% and 32% lower with Levemir (p<0.05) whereas the overall risk was similar between the groups. Hypoglycemia risk was estimated according to the incidence reported during the 20-week maintenance period. The 6-week titration period was excluded to eliminate any bias resulting from titration differences (glargine was a marketed product; Levemir was investigational). The incidence of nocturnal hypoglycemia with Levemir was estimated to be 4.3 episodes per subject-year and to be 6.6 episodes per subject-year for glargine (p<0.05). This represents a 32% reduction in nocturnal hypoglycemia risk with Levemir. Hypoglycemia was classified as follows: Major: Requiring assistance from another person Minor: Glucose measurement <56 mg/dL† Symptomatic only: No glucose measurement or glucose =56 mg/dL† Nocturnal: Any episode between 11 pm and 6 am †whole blood glucose converted to plasma equivalents Please see full prescribing information. Levemir® (insulin detemir [rDNA origin] injection) Novo Nordisk, Princeton, NJ. Reference: Pieber TR, Treichel H-C, Robertson LI, Mordhorst L, Gall M-A. Insulin detemir plus insulin aspart is associated with less risk of major as well as nocturnal hypoglycaemia than insulin glargine plus insulin aspart at comparable levels of glycaemic control in type 1 diabetes. Diabetologia. 2005;48(suppl 1):A92. The following is to be answered only in response to an unsolicited question. Levemir® demonstrated comparable A1c with less major and nocturnal hypoglycemia. The risk of major and nocturnal (11 pm to 6 am) hypoglycemia was 72% and 32% lower with Levemir (p<0.05) whereas the overall risk was similar between the groups. Hypoglycemia risk was estimated according to the incidence reported during the 20-week maintenance period. The 6-week titration period was excluded to eliminate any bias resulting from titration differences (glargine was a marketed product; Levemir was investigational). The incidence of nocturnal hypoglycemia with Levemir was estimated to be 4.3 episodes per subject-year and to be 6.6 episodes per subject-year for glargine (p<0.05). This represents a 32% reduction in nocturnal hypoglycemia risk with Levemir. Hypoglycemia was classified as follows: Major: Requiring assistance from another person Minor: Glucose measurement <56 mg/dL† Symptomatic only: No glucose measurement or glucose =56 mg/dL† Nocturnal: Any episode between 11 pm and 6 am †whole blood glucose converted to plasma equivalents Please see full prescribing information. Levemir® (insulin detemir [rDNA origin] injection) Novo Nordisk, Princeton, NJ. Reference: Pieber TR, Treichel H-C, Robertson LI, Mordhorst L, Gall M-A. Insulin detemir plus insulin aspart is associated with less risk of major as well as nocturnal hypoglycaemia than insulin glargine plus insulin aspart at comparable levels of glycaemic control in type 1 diabetes. Diabetologia. 2005;48(suppl 1):A92.

25. Detemir Is Associated With Less Weight Gain– Basal Insulin Trials: vs. NPH or Glargine Data from the Levemir treat-to-target trials can be compared to the Lantus treat-to-target studies. Data are presented as change in weight only. There is no indication of baseline weight (like on slide 20 where baseline HbA1c is shown), since baseline weight is not always reported: Riddle 2003 – not reported Hermansen 2006 – Det, 82.7kg NPH, 82.5 kg Philis-Tsimikas – not reported Data from the Levemir treat-to-target trials can be compared to the Lantus treat-to-target studies. Data are presented as change in weight only. There is no indication of baseline weight (like on slide 20 where baseline HbA1c is shown), since baseline weight is not always reported: Riddle 2003 – not reported Hermansen 2006 – Det, 82.7kg NPH, 82.5 kg Philis-Tsimikas – not reported

26. 6 Months Later…Taking Detemir 40 Units Daily

27. One Year Later… Still Taking Detemir Only

28. INITIATE Trial – Glargine vs. Premixed Insulin Analog: Study Design

29. INITIATE TrialPrimary End Point (A1C)

30. INITIATE TrialPatients Achieving A1C Goal 70/30 Glargine P-Value Baseline 9.7 ± 1.5 (n = 117) 9.8 ± 1.4 (n = 116) 0.4782 Study end 6.9 ± 1.1 (n = 110) 7.4 ± 1.3 (n = 114) 0.0010 Abs ? -2.79 ± 0.11 -2.36 ± 0.11 0.0057 70/30 Glargine P-Value Baseline 9.7 ± 1.5 (n = 117) 9.8 ± 1.4 (n = 116) 0.4782 Study end 6.9 ± 1.1 (n = 110) 7.4 ± 1.3 (n = 114) 0.0010 Abs ? -2.79 ± 0.11 -2.36 ± 0.11 0.0057

31. Considerations for PremixedInsulin Analogs Provides rapid- and intermediate-acting insulin in one injection Ability to intensify from a starting regimen of 1 injection of premixed insulin to 2 or 3 injections with same insulin Requires a relatively consistent meal and exercise pattern

33. PREFER Design: Premix Analog vs. All-Analog Basal-Bolus Therapy Design A 26-week, multinational, multicentre, treat-to-target trial. 719 subjects were randomised to BIAsp 30 BD or IDet OD/BD plus IAsp TID in a 1:3 ratio (by sealed code), respectively. All OADs were discontinued in both arms. BIAsp 30 consists of 30% soluble insulin aspart and 70% protamine-crystalised insulin aspart). Inclusion criteria Adults with type 2 diabetes =6 months diagnosed duration, BMI =40 kg/m2, currently receiving 1 or 2 OADs, with or without concomitant once-daily intermediate or long-acting insulin (NPH or insulin glargine). HbA1c at entry between 7.0% and 12.0%. Exclusion criteria Subjects treated with 3 or more OADs, or had been treated with short-acting or premixed insulin (regular or analogue) during the previous 6 months. Subjects with proliferative retinopathy or maculopathy laser-treated within 6 months or requiring immediate treatment, with impaired hepatic or renal function (alanine aminotransferase =2x upper reference limit or creatinine =150 µmol/l, respectively), or with cardiac disease (unstable angina, myocardial infarction within 6 months or congestive heart failure NYHA III or IV) were also excluded. After randomisation, the first 6 weeks constituted the titration phase, during which doses were adjusted weekly according to defined algorithms; optimization was continued after the titration period if control targets had not been achieved. Detemir titrated to pre-breakfast 4-7mmol/L, Asp to achieve 90min postprandial glucose < 10mmol/L; 2nd dose added if pre-dinner > 7.0mmol/L despite at-target pre-breakfast and optimised lunchtime Asp dose Design A 26-week, multinational, multicentre, treat-to-target trial. 719 subjects were randomised to BIAsp 30 BD or IDet OD/BD plus IAsp TID in a 1:3 ratio (by sealed code), respectively. All OADs were discontinued in both arms. BIAsp 30 consists of 30% soluble insulin aspart and 70% protamine-crystalised insulin aspart). Inclusion criteria Adults with type 2 diabetes =6 months diagnosed duration, BMI =40 kg/m2, currently receiving 1 or 2 OADs, with or without concomitant once-daily intermediate or long-acting insulin (NPH or insulin glargine). HbA1c at entry between 7.0% and 12.0%. Exclusion criteria Subjects treated with 3 or more OADs, or had been treated with short-acting or premixed insulin (regular or analogue) during the previous 6 months. Subjects with proliferative retinopathy or maculopathy laser-treated within 6 months or requiring immediate treatment, with impaired hepatic or renal function (alanine aminotransferase =2x upper reference limit or creatinine =150 µmol/l, respectively), or with cardiac disease (unstable angina, myocardial infarction within 6 months or congestive heart failure NYHA III or IV) were also excluded. After randomisation, the first 6 weeks constituted the titration phase, during which doses were adjusted weekly according to defined algorithms; optimization was continued after the titration period if control targets had not been achieved. Detemir titrated to pre-breakfast 4-7mmol/L, Asp to achieve 90min postprandial glucose < 10mmol/L; 2nd dose added if pre-dinner > 7.0mmol/L despite at-target pre-breakfast and optimised lunchtime Asp dose

34. Similar A1C Reduction In Insulin-Naïve But Not Insulin-Treated Patients HbA1c and previous insulin therapy Patients previously treated with insulin had significantly greater HbA1c reductions with IDet/IAsp than with BIAsp 30 (1.21% vs. 0.75%; p = 0.0129), whereas insulin-naïve patients had similar reductions with both regimens (1.69% vs. 1.42%; p = 0.106). HbA1c reductions were greater for both treatments in insulin-naïve subjects compared with subjects previously treated with basal insulin. In the IDet/IAsp group, achieved HbA1c levels were similar in patients transferred to twice-daily IDet and in the majority who remained on once-daily IDet (6.97% vs. 6.93%; p = 0.697).HbA1c and previous insulin therapy Patients previously treated with insulin had significantly greater HbA1c reductions with IDet/IAsp than with BIAsp 30 (1.21% vs. 0.75%; p = 0.0129), whereas insulin-naïve patients had similar reductions with both regimens (1.69% vs. 1.42%; p = 0.106). HbA1c reductions were greater for both treatments in insulin-naïve subjects compared with subjects previously treated with basal insulin. In the IDet/IAsp group, achieved HbA1c levels were similar in patients transferred to twice-daily IDet and in the majority who remained on once-daily IDet (6.97% vs. 6.93%; p = 0.697).

35. Optimization of Intensive Insulin Therapy Basal requirements Prandial requirements “Correction dose” Correction based on premeal glucose level to try to achieve optimal glucose level at next meal Increase prandial insulin if premeal glucose too high Decrease prandial insulin if premeal glucose too low

36. Optimization of Intensive Insulin Therapy Prandial insulin dose based on Amount of carbohydrate consumed (advanced carbohydrate counting) Premeal glucose Sensitivity factor– 1 unit of prandial insulin will lower glucose by x mg/dL in basal state

37. Conversion to Carbohydrate Counting: Dose Calculation The “450 Rule” for the mealtime ratios Let’s assume patient is taking 70 units of insulin per day Divide 450 by the total daily dose (TDD) of insulin 450/70 = 6.4, rounded off to 6 grams. Thus, the starting ratio is 1:6 (1 unit rapid acting insulin for every 6 grams of carbohydrate eaten). Can adjust upward (breakfast) or downward (lunch, supper) Segment 15: Case 2 Continued (Speaker 2) Segment time allocation: 6 minutes, Estimated run time: 8:06 – 8:12 PM Comment: Basal/bolus (Physiologic) insulin initiation dosing information, step two. Describes the methods of calculating the ratio’s by the “450 rule” and the sensitivity or correction factor by the “1500 rule.” Note that this is the way we approach this at Joslin, but it is not the only way that is used. Some educators and endos use other "rules", like the 1700 rule or the 500 rule.  Our Joslin policy is to use 1500 and 450, mostly because there isn't a lot of science behind any of this, and also, that in most cases, the ratios and Sensitivity Factors need to be adjusted anyway. Please present this version, but then if you wish to present others, you may do so briefly. Segment 15: Case 2 Continued (Speaker 2) Segment time allocation: 6 minutes, Estimated run time: 8:06 – 8:12 PM Comment: Basal/bolus (Physiologic) insulin initiation dosing information, step two. Describes the methods of calculating the ratio’s by the “450 rule” and the sensitivity or correction factor by the “1500 rule.” Note that this is the way we approach this at Joslin, but it is not the only way that is used. Some educators and endos use other "rules", like the 1700 rule or the 500 rule.  Our Joslin policy is to use 1500 and 450, mostly because there isn't a lot of science behind any of this, and also, that in most cases, the ratios and Sensitivity Factors need to be adjusted anyway. Please present this version, but then if you wish to present others, you may do so briefly.

38. The “1500 Rule” for the Sensitivity Factor Divide 1500 by the total daily dose (TDD) of insulin If total daily dose of insulin is 70 units 1500/70 = 21, rounded off to 20 units. Thus, figure that 1 unit of rapid acting insulin will lower the glucose 20 mg/dL

39. In Summary Type 2 diabetes is a progressive disease characterized by increasing ß-cell dysfunction Many patients require insulin to achieve optimal control Initiation and intensification of insulin treatment is based on an understanding of the normal physiology of insulin secretion and can be successfully achieved using basal and prandial insulins

40. Breakout: Initiation and Titration of Insulin Martin J. Abrahamson, MD Medical Director and Senior Vice President Joslin Diabetes Center Associate Professor of Medicine Harvard Medical School Boston, Massachusetts Martha M. Funnell, MS, RN, CDE Michigan Diabetes Research and Training Center Juvenile Diabetes Research Foundation Center for the Study of the Complications in Diabetes Ann Arbor, Michigan