Clinical Care/Education/Nutrition ORIGINAL ARTICLE

Continuous Subcutaneous Infusion (CSII) of Insulin Aspart Versus Multiple Daily Injection of Insulin Aspart/Insulin Glargine in Type 1 Diabetic Patients Previously Treated With CSII

1 6 IRL B. HIRSCH, MD PETER HU, PHD ultiple daily injection (MDI) ther- 2 6 BRUCE W. BODE, MD OLGA M. SANTIAGO, MD apy and continuous subcutane- 3 7 SATISH GARG, MD JERZY W. KOLACZYNSKI, MD 4 ous insulin infusion (CSII) with ENDY ANE MD FOR THE NSULIN SPART M W S. L , I A CSII/MDI 5 an external pump are two current meth- ALLEN SUSSMAN, MD COMPARISON STUDY GROUP ods of intensive insulin therapy for diabe- tes. MDI therapy requires bolus injection of short- or rapid-acting insulin at each OBJECTIVE — Multiple daily injection (MDI) therapy of bolus insulin aspart and basal meal, along with long-acting insulin once insulin glargine was compared with continuous subcutaneous insulin infusion (CSII) with aspart or twice daily for basal insulin coverage. in type 1 diabetic patients previously treated with CSII. Rapid-acting insulin analogs are adminis- tered as mealtime boluses to control post- RESEARCH DESIGN AND METHODS — One hundred patients were enrolled in a prandial glycemic excursions and have randomized, multicenter, open-label, crossover study. After a 1-week run-in period with aspart been shown to be more effective than reg- by CSII, 50 subjects were randomly assigned to MDI therapy (aspart immediately before each meal and glargine at bedtime) and 50 subjects continued CSII. After 5 weeks of the first treat- ular human insulin (1,2). The long-acting ment, subjects crossed over to the alternate treatment for 5 weeks. During the last week of each insulin analog insulin glargine has pro- treatment period, subjects wore a continuous glucose monitoring system for 48–72 h. longed pharmacodynamics that make it suitable for use as a basal insulin (3,4). RESULTS — Mean serum levels were significantly lower after CSII therapy than The growing use of CSII therapy is after MDI therapy (343 Ϯ 47 vs. 355 Ϯ 50 ␮mol/l, respectively; P ϭ 0.0001). Continuous based on its proven efficacy, improve- glucose monitoring profiles over a 24-h time period showed that glucose exposure was 24 and ments in pump technology, and increased 40% lower for CSII than MDI as measured by area under the curve (AUC) glucose Ն80 mg/dl Ϫ patient preference (5–10). Occasionally, (1,270 Ϯ 742 vs. 1,664 Ϯ 1,039 mg h dl 1; P Ͻ 0.001) and AUC glucose Ն140 mg/dl (464 Ϯ Ϫ1 patients utilizing CSII therapy may have 452 vs. 777 Ϯ 746 mg h dl , CSII vs. MDI, respectively; P Ͻ 0.001). Similar percentages of subjects reported hypoglycemic episodes (CSII: 92%, MDI: 94%) and nocturnal (12:00 A.M.to to temporarily discontinue CSII use be- 8:00 A.M.) hypoglycemic episodes (CSII: 73%, MDI: 72%). Major was infrequent cause of pump malfunction, skin prob- (CSII: two episodes, MDI: five episodes). lems, or physical activity (especially contact sports and water sports). During CONCLUSIONS — In a trial of short duration, CSII therapy with insulin aspart resulted in such periods, type 1 diabetic patients lower glycemic exposure without increased risk of hypoglycemia, as compared with MDI with switching to MDI therapy could continue insulin aspart and glargine. to use insulin aspart as the mealtime in- Care 28:533–538, 2005 sulin and could use insulin glargine as the ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● basal insulin. Previous studies have shown that From the 1University of Washington, Seattle, Washington; the 2Atlanta Diabetes Associates, Atlanta, Georgia; the 3Barbara Davis Center for Childhood Diabetes, Denver, Colorado; the 4Mountain Diabetes and Endo- CSII is at least equivalent to, and often crine Center, Asheville, North Carolina; the 5Rainier Clinical Research Center, Renton, Washington; 6Novo more effective than, MDI therapy (10– Nordisk Pharmaceuticals, Princeton, New Jersey; and the 7University of Medicine and Dentistry of New 12). The use of an analog-only MDI regi- Jersey, Robert Wood Johnson Medical School, New Brunswick, New Jersey. men consisting of basal glargine and Address correspondence and reprint requests to Irl B. Hirsch, MD, University of Washington Medical Center, 1959 NE Pacific St., Box 356176, Seattle, WA 98195. E-mail: [email protected]. mealtime rapid-acting analog has been Received for publication 2 May 2004 and accepted in revised form 20 November 2004. nicknamed “poor man’s pump.” How- I.R.H. is a paid consultant for Aventis, Eli Lilly, Medtronic, and Novo Nordisk and has received grants ever, its safety and efficacy compared with from Novo Nordisk and Pfizer; B.W.B. is a paid consultant for Medtronic, Novo Nordisk, and Aventis and CSII has never been tested in a controlled, has received grants from Novo Nordisk, Aventis, and Medtronic; S.G. has received grants from Novo cross-over clinical trial setting. Conse- Nordisk; and W.L. has received honoraria from Aventis and Novo Nordisk. Abbreviations: AUC, area under the curve; CGMS, continuous glucose monitoring system; CSII, con- quently, we compared insulin aspart as tinuous subcutaneous insulin infusion; MDI, multiple daily injection; SMBG, self-measured blood glucose. the only approved rapid-acting analog for A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion CSII with insulin aspart/glargine in MDI factors for many substances. therapy. Because the principles applied to © 2005 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby mealtime coverage were identical during marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. both treatments, the study tests whether

DIABETES CARE, VOLUME 28, NUMBER 3, MARCH 2005 533 CSII versus MDI in the theoretically smooth profile of unit-for-unit basis, to MDI therapy using CGMS data. The sensor failure rate was glargine will provide safety and efficacy a single basal bedtime injection of insulin ϳ20%. Glucose exposures were com- comparable to the “customized” basal in- glargine (Lantus, Aventis Pharmaceuti- pared between treatment groups and sulin delivery that can be programmed by cals), consistent with product labeling in were based on the first 24 h of CGMS an infusion pump during CSII. the U.S. at that time, and mealtime bo- monitoring. The 24-h glycemic profiles luses of insulin aspart delivered by the were calculated from the area under the RESEARCH DESIGN AND Induo device. Subjects assigned to CSII curve (AUC) for glucose values Ն80 METHODS — This was a multicenter, treatment during the study used their mg/dl (4.4 mmol/l) and Ն140 mg/dl (7.8 open-label, randomized, two-period, own prestudy insulin pumps. mmol/l). CGMS profiles had to be at least crossover, 10-week study in which CSII In both groups, the basal insulin cov- 24 h in duration to be suitable for use in was compared with MDI therapy in two erage could be modified through the sched- the AUC calculations. For the CGMS data 5-week treatment periods. The study was uled phone communications with the analysis, glycemia of 80 mg/dl was the conducted at 15 centers in the U.S., in subject during a 1-week dose-adjustment minimal value considered optimal. Ac- accordance with the Declaration of Hel- period with a goal of fasting and pre- cordingly, glycemic control was consid- sinki and good clinical practice guidelines dinner glycemia in the range of 90–126 ered to be better/tighter when the glucose (13). All subjects provided written in- mg/dl (5–7 mmol/l). The investigator exposure of AUC Ն80 mg/dl was mini- formed consent. Analyses of blood sam- made final adjustments to the basal insu- mized in a treatment regimen. ples for safety and efficacy parameters lin dose at the visit scheduled for the end were performed by Medical Research Lab- of the 1-week dose-adjustment period, af- Safety assessments oratories International (Highland ter which the adjusted basal insulin rate/ Safety assessments included adverse Heights, KY). dose was maintained for the remaining 4 events, physical examination findings, Enrolled type 1 diabetic subjects were weeks of the treatment period. At the end and clinical laboratory evaluations. Ad- Ն18 years old, had BMI Յ40 kg/m2, and of the 5-week treatment period, subjects verse events were recorded throughout Յ had HbA1c 9%. All were previously crossed over to the alternate treatment ther- the study, and a general physical exami- treated by CSII for at least 3 months be- apy (e.g., CSII to MDI), with a 1-week nation was conducted at the beginning fore the screening visit. Subjects with im- basal dose adjustment period immedi- and end of the study. paired hepatic or renal function (alanine ately followed by 4 weeks of treatment. The frequency of hypoglycemic epi- aminotransferase or creatinine values Ն2 sodes was monitored. Minor hypogly- times the upper limit of the normal refer- Efficacy assessments cemic episodes were defined as any ence range for the age-group), impaired Overall glycemic control was assessed by asymptomatic blood glucose measure- cardiac function, hypoglycemia unaware- fructosamine measurements taken at the ment Ͻ50 mg/dl, or as episodes with ness, or recurrent major hypoglycemia beginning and end of each treatment pe- symptoms consistent with hypoglycemia were excluded. Subjects performed self- riod (normal fructosamine range: 0–285 with confirmation by blood glucose mea- ␮ Ͻ measured blood glucose (SMBG) mea- mol/l). HbA1c values were measured at surement 50 mg/dl that were handled surements and were knowledgeable in the the beginning and end of the study (nor- by the subject. Major hypoglycemic epi- methods of prandial insulin bolus dose mal HbA1c range: 4–6%). Overall glucose sodes were defined as episodes with se- adjustment based on preprandial glyce- control and postprandial glycemic con- vere central nervous system symptoms mia and carbohydrate intake. Women of trol were assessed by comparison of consistent with hypoglycemia that the pa- child-bearing age were excluded if they 8-point SMBG profiles (blood glucose tient was unable to treat himself/herself, were pregnant, breast-feeding, or not readings before and 2 h after breakfast, which had either 1) blood glucose Ͻ50 practicing contraception. Subjects and in- lunch, and dinner and at 12:00 A.M. and mg/dl or 2) reversal of symptoms after ei- vestigators were blinded to treatment se- 3:00 A.M.) performed at the end of each ther food intake or glucagon/intravenous quence up to the point of subject treatment period. glucose administration. Symptomatic hy- randomization. During the last week of each 5-week poglycemic episodes were symptoms that For a 1-week period before random- treatment period, subjects wore a contin- were considered to be related to hypogly- ization, subjects were switched from use uous glucose monitoring system (CGMS) cemia but not confirmed by blood glucose of their prestudy CSII insulin to insulin (MiniMed MMT-7102; Medtronic, measurement Ͻ50 mg/dl. aspart on a unit-by-unit basis. Subjects Northridge, CA) for up to 72 h. Data were Safety assessments included hematol- were instructed on insulin dose adjust- obtained from the CGMS using Solutions ogy (red blood cell, white blood cell, he- ment and on the use of the Induo device 7314 Sensor Data Export Utility software matocrit, and hemoglobin) and blood (LifeScan, Milpitas, CA). In general, the (Medtronic). MiniMed Glucose Sensors chemistry parameters (creatinine, total mealtime insulin coverage with aspart bo- (MMT-7002; Medtronic) were used. Glu- protein, liver function tests, lactic dehy- luses during MDI and CSII treatments fol- cose sensor signals were acquired every drogenase, sodium, and potassium). lowed the same principles of dose 10 s, and the average over 5 min was adjustments based on carbohydrate saved in memory. Accuracy of glucose Statistical analysis counting and a preprandial blood glucose sensing was ensured by using CGMS data Fructosamine and AUC glucose parame- value. At randomization, subjects either that had a correlation of at least 0.79 with ters based on 24-h glycemic profile from remained on CSII with insulin aspart (No- standard blood glucose meter readings. CGMS were analyzed using a standard voLog, Novo Nordisk Pharmaceuticals, Thirty-six subjects had to have their glu- ANCOVA model. CGMS profiles had to Princeton, NJ) or were switched, on a cose sensor reimplanted to obtain usable be at least 24 h in duration to be used in

534 DIABETES CARE, VOLUME 28, NUMBER 3, MARCH 2005 Hirsch and Associates

Table 1—Characteristics of enrolled population and rate of study completion was significantly less than their baseline value (7.2 Ϯ 0.7 vs. 7.5 Ϯ 0.8%, respec- Treatment sequence tively; P Ͻ 0.01). CSII to MDI MDI to CSII All subjects Based on subjects with available CGMS data, the AUC glucose values Ն80 n 50 50 100 and Ն140 mg/dl were significantly re- Age (years) 41.7 Ϯ 11.1 44.2 Ϯ 11.0 43.0 Ϯ 11.1 duced for subjects during CSII treatment Sex (%) (Table 2) compared with MDI treatment. Male 19 (38) 18 (36) 37 (37) The observation of reduced AUC glucose Female 31 (62) 32 (64) 63 (63) values for CSII treatment was not depen- Race (%) dent on treatment sequence. During the Caucasian 48 (96) 48 (96) 96 (96) CGMS monitoring period, CSII-treated Hispanic 2 (4) 1 (2) 3 (3) subjects spent significantly more time in Asian 0 (0) 1 (2) 1 (1) the glucose range Ն80 but Յ140 mg/dl BMI (kg/m2) 27.1 Ϯ 4.1 26.7 Ϯ 4.0 26.9 Ϯ 4.0 than MDI-treated subjects, as demon- Ϯ Ϯ Ϯ HbA1c at screening (%) 7.5 0.8 7.4 0.8 7.5 0.8 strated by the greater percentage of glyce- Duration of diabetes (years) 19.7 Ϯ 11.3 23.9 Ϯ 12.3 21.8 Ϯ 11.9 mic measurements in that range (43 vs. Daily insulin dose (IU/kg) 0.54 Ϯ 0.23 0.54 Ϯ 0.21 0.54 Ϯ 0.22 33% of readings for the CSII vs. MDI sub- (n ϭ 45) (n ϭ 50) (n ϭ 95) jects, respectively; P Ͻ 0.0001). CSII- Basal 0.27 Ϯ 0.10 0.29 Ϯ 0.14 0.28 Ϯ 0.12 treated subjects also spent significantly Bolus 0.29 Ϯ 0.18 0.25 Ϯ 0.11 0.27 Ϯ 0.15 less time in the glucose range Ͼ140 mg/dl Daily basal pump rates (%) than MDI-treated subjects (41 vs. 50% of One basal rate 12 (24) 6 (12) 18 (18) readings for the CSII vs. MDI subjects, Two basal rates 6 (12) 7 (14) 13 (13) respectively; P Ͻ 0.0001). The percent- Three basal rates 14 (28) 14 (28) 28 (28) age of glycemic readings Ͻ80 mg/dl was Four basal rates 18 (36) 23 (46) 41 (41) similar for both treatments (17% of all Subjects completing study (%) 45 (90) 46 (92) 91 (91) readings). Withdrawals during treatment (%) 5 (10) 4 (8) 9 (9) The mean blood glucose profiles from Adverse events (n) 000the CGMS readings during the 1st com- Noncompliance (%) 2 (4) 3 (6) 5 (5) plete day of monitoring (12:00 A.M.to Ineffective therapy (%) 1 (2) 1 (2) 2 (2) 12:00 A.M.) are presented in Fig. 1. The Withdrew consent (%) 2 (4) 0 2 (2) generally lower blood glucose values dur- Data are means Ϯ SD or n (%). Treatment sequence refers to the order of insulin administration in the two ing evening, nighttime, and morning for treatment periods (i.e., CSII to MDI refers to CSII treatment during period 1 and MDI treatment during the CSII group represent those time peri- period 2). ods that contribute to the lower overall AUC glucose values determined for the the analysis. For analysis of end-of-period SMBG profiles for combined subjects at CSII group from the CGMS profiles. fructosamine values, the last observation the end of the CSII or MDI treatment pe- Subjects in both treatment sequences carried forward approach was used. riods showed that the overall blood glu- reported using total daily insulin doses cose profiles were not significantly during CSII and MDI treatments that RESULTS — Demographic variables different between treatments (P ϭ 0.074). were similar to their baseline total daily of age, HbA1c, and BMI were similar at Although not different, the blood glucose insulin dose (Table 2). For both treatment baseline for subjects in both treatment se- values after breakfast (CSII 158 Ϯ 63 regimens, the total daily insulin dose for quences (Table 1). Prior insulin use was mg/dl and MDI 182 Ϯ 82 mg/dl), before combined subjects was divided nearly balanced between treatment sequences: dinner (128 Ϯ 58 and 148 Ϯ 71 mg/dl), equally between basal and bolus doses in each sequence, 41, 7, and 2 subjects and after dinner (144 Ϯ 64 and 159 Ϯ 77 (22.4 Ϯ 9.6/21.3 Ϯ 12.7 vs. 23.6 Ϯ 10.9/ used insulin lispro, insulin aspart, and mg/dl) tended to be lower for CSII ther- 22.5 Ϯ 11.1 units for basal/bolus doses in buffered regular insulin, respectively. apy than MDI therapy. Results of the CSII vs. MDI groups, respectively). 8-point SMBG profiles were not affected Efficacy by treatment sequence. Safety Overall, CSII therapy with insulin aspart Subjects maintained overall glycemic The numbers of adverse events and sub- provided improved glycemic control control during the study. The mean jects reporting those events were similar compared with MDI therapy with insulin HbA1c values for subjects at the end of the between treatment therapies. One subject aspart/glargine as measured by fruc- 10-week study were similar between reported moderate tosamine values and 8-point SMBG pro- treatment sequences (CSII to MDI 7.3 Ϯ during CSII treatment; the event was re- files. The mean fructosamine value 0.7% vs. MDI to CSII 7.1 Ϯ 0.7%, P Ͼ solved on the same day. There were no determined at the end of the CSII treat- 0.05) and demonstrated that subjects reports of diabetic ketoacidosis in the ment for combined subjects was signifi- maintained overall glycemic control in MDI group. cantly lower than after treatment with both treatment sequences. The end-of- Hypoglycemic episodes were experi- MDI therapy (Table 2). The mean 8-point study HbA1c value for combined subjects enced by 92% of the subjects during CSII

DIABETES CARE, VOLUME 28, NUMBER 3, MARCH 2005 535 CSII versus MDI in type 1 diabetes

Table 2—Glycemic parameters at the end of each treatment period

Treatment Glycemic parameter Difference Treatment sequence n Baseline n CSII n MDI (CSII Ϫ MDI) P Mean fructosamine (␮mol/l) CSII to MDI 50 351 Ϯ 44 48 352 Ϯ 46 49 360 Ϯ 49 MDI to CSII 49 345 Ϯ 48 50 334 Ϯ 48 48 349 Ϯ 50 Combined subjects NA 98* 343 Ϯ 47 97 355 Ϯ 50 Ϫ11.8 [Ϫ13.4 to Ϫ4.63] 0.0001† Mean AUC glucose Ն80 mg/dl (mb h dlϪ1)‡ CSII to MDI NA 37 1,150 Ϯ 758 34 1,605 Ϯ 1018 MDI to CSII NA 33 1,403 Ϯ 713 37 1,718 Ϯ 1,069 Combined subjects NA 70 1,270 Ϯ 742 71 1,664 Ϯ 1039 Ϫ394 [Ϫ654 to Ϫ196] 0.0005† Mean AUC glucose Ն140 mg/dl (mg h dlϪ1)‡ CSII to MDI NA 37 381 Ϯ 397 34 774 Ϯ 721 MDI to CSII NA 33 557 Ϯ 496 37 779 Ϯ 778 Combined subjects NA 70 464 Ϯ 452 71 777 Ϯ 746 Ϫ313 [Ϫ489 to Ϫ150] 0.0004† Daily insulin dose (units)§ CSII to MDI 45 42.3 Ϯ 17.9 50 42.1 Ϯ 19.2 46 46.0 Ϯ 18.2 MDI to CSII 50 41.6 Ϯ 16.1 48 39.6 Ϯ 17.5 50 46.2 Ϯ 20.5 Combined subjects 95 41.9 Ϯ 16.9 98 40.9 Ϯ 18.4 96 46.1 Ϯ 19.4 0.08 Data are means Ϯ SD or difference [95% CI]. *Numbers of subjects are greater than those who completed the trial because fructosamine was determined at an end- of-study visit; the last observation carried forward approach was used for fructosamine; †statistical inference was made for combined values only; treatment, sequence, and center were included as fixed effect, subject as random effect in the model; ‡AUC glucose was based on the first 24 h of CGMS monitoring; §treatment insulin dose values were measured after the 1-week dose adjustment period for each treatment period. treatment and by 91% of the subjects CONCLUSIONS — The 5-week cross- apy in the late postprandial period or by during MDI therapy (Table 3). Fifty-three over treatment periods in this clinical trial the waning effect of insulin glargine at din- percent of all hypoglycemic episodes dur- were not of sufficient duration to use nertime during MDI therapy. In a study ing both treatments were symptomatic HbA1c as a primary efficacy end point. switching type 1 diabetic patients from hypoglycemic episodes (not confirmed However, the significantly lower fruc- twice-daily NPH insulin during MDI to with blood glucose Ͻ50 mg/dl). The rates tosamine values and significantly reduced once-daily glargine during MDI, one- of daily minor hypoglycemic episodes AUC glucose values during CSII therapy quarter of the subjects required further were not different between treatments. indicate that CSII therapy with insulin as- switching to twice-daily glargine injec- However, the rate of nocturnal minor hy- part provides better glycemic control than tions to achieve acceptable glycemic con- poglycemic episodes was significantly less MDI therapy with insulin aspart/insulin trol (14). The switch to twice-daily glargine for subjects treated with CSII than for glargine. was prompted by an increase in HbA1c or subjects treated with MDI therapy. In The 24-h CGMS blood glucose pro- persistent elevation of the predinner blood contrast, the rate of daytime minor hypo- files demonstrate that CSII therapy pro- glucose, despite efforts to titrate both bolus glycemic episodes was significantly vided improved glycemic control during insulin and glargine insulin. In the present greater for subjects treated with CSII than the nighttime and morning hours (Fig. 1), trial, the tested drug products were used for subjects treated with MDI therapy (Ta- time periods that contributed substan- according to their existing U.S. labeling. At ble 3). Seven major hypoglycemic epi- tially to lowering the AUC glucose param- the time of the study, glargine was indicated sodes occurred during the study: two eters Ͼ80 and Ͼ140 mg/dl for the CSII for single-dose bedtime injections. We CSII-treated subjects had a single epi- group. chose not to explore off-label use during the sode, two MDI-treated subjects had two CSII allows the nighttime basal insu- trial. The 1-week titration phase may also episodes each, and one MDI-treated pa- lin rate to be fine-tuned; thus, it may pro- have been too short for optimization of tient had a single episode. No subjects vide a distinct advantage over MDI glargine in the MDI regimen. had a major hypoglycemic episode during therapy by being better able to control the The improvement in glycemic control both treatments. dawn phenomenon and possibly curtail for the CSII group did not come at the No end-of-study differences in blood the exacerbation of postprandial hyper- expense of an increase in insulin dose chemistry or hematology laboratory val- glycemia at breakfast. The reduced post- (Table 2). However, the trend was for a ues were noted for the study population. prandial rise after dinner with CSII, slightly lower total daily insulin dose in Mean values for vital signs and weight at despite standardized dinnertime bolus in- the CSII group regardless of treatment se- the end of the study were similar to base- sulin dosing, might be explained by better quence, although the difference was not line values. adjustment of the basal rate by CSII ther- significant.

536 DIABETES CARE, VOLUME 28, NUMBER 3, MARCH 2005 Hirsch and Associates

Figure 1—Mean blood glucose profiles from continuous glucose monitoring. The 24-h profile (12:00 A.M. to 12:00 A.M.) is the mean blood glucose profile for the 1st complete day for combined subjects during the last week of each treatment. Only the 1st 24-h (12:00 A.M. to 12:00 A.M.) profile was presented because the alignment of CGMS profiles into proper time shift limited the available data for a profile containing the 2nd 24 h. Each time point represents the mean of all available data that had a correlation of at least 0.79 with standard blood glucose readings. Number of subjects per .MDI (E), 54–57 ;57–50 ,(ء) time point for each profile: CSII

Subjects were able to switch between ences in glycemic control and risk of ing therapy from CSII with insulin aspart CSII and MDI therapy without increasing hypoglycemia events were small but sta- to MDI with insulin aspart/insulin the overall rate of minor hypoglycemic tistically significant. Accordingly, switch- glargine was a clinically viable choice for episodes (6.2 vs. 5.7 episodes per subject during the 5-week treatment period, re- spectively). The significantly lower risk of Table 3 —Hypoglycemia reported at the end of each treatment period reported nocturnal hypoglycemia for CSII with insulin aspart is a noteworthy advan- CSII MDI tage over MDI using aspart/glargine. Subjects Episodes Rate* Subjects Episodes Rate* P† Avoiding nocturnal hypoglycemia is par- Symptomatic, minor, and ticularly important because it occurs major hypoglycemia when a patient may be unable to deal with Daily hypoglycemia 90 (92) 956 10.6 89 (91) 825 9.3 0.0041 the episode quickly and effectively. The Daytime hypoglycemia 87 (89) 736 8.5 87 (89) 541 6.2 Ͻ0.001 decreased risk of nocturnal hypoglycemia Nocturnal hypoglycemia 71 (72) 216 3.0 70 (73) 280 4.0 0.0024 during CSII treatment was offset by an Minor hypoglycemia increase in the risk of daytime minor hy- Daily hypoglycemia 72 (74) 447 6.2 68 (69) 387 5.7 0.2099 poglycemia, although the rate was rela- Daytime hypoglycemia 59 (60) 333 5.6 59 (60) 232 3.9 Ͻ0.001 tively low during both treatments (5.6 vs. Nocturnal hypoglycemia 51 (52) 110 2.2 49 (50) 155 3.2 0.0020 3.9 episodes per subject during the Symptomatic hypoglycemia 5-week treatment period for CSII vs. MDI Daily hypoglycemia 73 (75) 507 6.9 71 (72) 434 6.1 0.0506 Daytime hypoglycemia 70 (71) 403 5.8 64 (65) 305 4.8 0.0124 therapies, respectively). Nocturnal hypoglycemia 41 (42) 104 2.5 47 (48) 125 2.7 0.7211 When subjects temporarily inter- rupted CSII therapy using insulin aspart Data are n (%) or n unless otherwise indicated. Nocturnal episodes occurred between midnight and 8:00 A.M.; daytime episodes occurred between 8:00 A.M. and midnight. *Rate is based on the number of episodes by and used MDI therapy with insulin as- subjects having episodes during the 5-week treatment period; †calculated for the rate of hypoglycemia using part/insulin glargine, the observed differ- Poisson regression.

DIABETES CARE, VOLUME 28, NUMBER 3, MARCH 2005 537 CSII versus MDI in type 1 diabetes temporary glycemic management of type glargine (HOE901) and NPH human in- sion therapy in patients with type 1 dia- 1 diabetes. sulin used with insulin lispro in patients betes. Diabetes Care 24:69–72, 2001 In conclusion, this clinical trial of rel- with type 1 diabetes. Diabetes Care 23: 10. Tsui E, Barnie A, Ross S, Parkes R, and atively short duration indicates that CSII 1666–1671, 2000 Zinmann B: Intensive insulin therapy was a more optimal therapy than MDI, re- 4. Lepore M, Pampanelli S, Fanelli C, Por- with insulin lispro: a randomized trial of cellati F, Bartocci L, Di Vincenzo A, Cor- continuous subcutaneous insulin infu- sulting in lower glycemic exposure with- doni C, Costa E, Brunetti P, Bolli GP: sion versus multiple daily insulin injec- out an increased risk of hypoglycemia. Pharmacokinetics and pharmacodynam- tion. Diabetes Care 24:1722–1727, 2001 ics of subcutaneous injection of long-act- 11. Hanaire-Broutin H, Melki V, Bessie`res-La- ing human insulin analog glargine, NPH combe S, Tauber J-P, Study Group for the Acknowledgments— This clinical trial re- insulin, and ultralente human insulin and Development of Pump Therapy in Diabe- ceived financial support from Novo Nordisk. continuous subcutaneous infusion of in- tes: Comparison of continuous subcuta- We thank the following members of the In- sulin lispro. Diabetes 49:2142–2148, 2000 neous insulin infusion and multiple daily sulin Aspart CSII/MDI Comparison Study 5. Pickup J, Keen H: Continuous subcutane- injection regimens using insulin lispro Group: Samuel E. Crockett, Steven Edelman, ous insulin infusion at 25 years: evidence in type 1 diabetic patients on intensified Robert Evans, Norman Fishman, Brent Gooch, base for the expanding use of insulin treatment. Diabetes Care 23:1232–1235, Robert Henry, Rajeev Jain, Sanford N. Plevin, pump therapy in type 1 diabetes. Diabetes 2000 Mark Warren, and Richard Weinstein. Care 25:593–598, 1996 12. DeVries JH, Snoek FJ, Kostense PJ, Ma- 6. Reynolds RL: Reemergence of insulin surel N, Heine RJ, the Dutch Insulin pump therapy in the 1990s. South Med J Pump Study Group: A randomized trial of References 93:1157–1161, 2000 continuous subcutaneous insulin infu- 1. Lindholm A, McEwen J, Riis AP: Im- 7. Bode BW, Steed RD, Davidson PC: Reduc- sion and intensive injection therapy in proved postprandial glycemic control with tion in severe hypoglycemia with long- type 1 diabetes for patients with long- insulin aspart: a randomized double- term continuous subcutaneous insulin standing poor glycemic control. Diabetes blind cross-over trial in type 1 diabetes. infusion in type 1 diabetes. Diabetes Care Care 25:2074–2080, 2002 Diabetes Care 22:801–805, 1999 19:324–327, 1996 13. World Medical Association declaration 2. Anderson JH, Brunelle RL, Koivosto VA, 8. Bode BW, Weinstein R, Bell D, McGill J, of Helsinki: recommendations guiding Pfutzner A, Trautmann ME, Vignati L, Nadeau D, Raskin P, Davidson J, Henry R, physicians in biomedical research in- DiMarchi R, Multicenter Insulin Lispro Huang W-C, Reinhardt RR: Comparison volving human subjects. JAMA 277:925– Study Group: Reduction of postprandial of insulin aspart with buffered regular in- 926, 1997 in IDDM patients on insu- sulin and insulin lispro in continuous 14. Albright ES, Desmond R, Bell D: Efficacy lin-analog treatment. Diabetes 46:265–270, subcutaneous insulin infusion. Diabetes of conversion from bedtime NPH insulin 1997 Care 25:439–444, 2002 injection to once- or twice-daily injec- 3. Raskin P, Klaff L, Bergenstal R, Halle JP, 9. Bode BW, Strange P: Efficacy, safety, and tions of insulin glargine in type 1 diabetic Donley D, Mecca T: A 16-week compari- pump compatibility of insulin aspart used patients using basal/bolus therapy. Diabe- son of the novel insulin analog insulin in continuous subcutaneous insulin infu- tes Care 27:632–633, 2004

538 DIABETES CARE, VOLUME 28, NUMBER 3, MARCH 2005