- Email: [email protected]
Risk for nocturnal hypoglycemia with biphasic insulin aspart 30 compared with biphasic human insulin 30 in adults with type 2 diabetes mellitus: A meta-analysis
Clinical Therapeutics/Volume 31, Number 8, 2009
Risk for Nocturnal Hypoglycemia With Biphasic Insulin Aspart 30 Compared With Biphasic Human Insulin 30 in Adults With Type 2 Diabetes Mellitus: A Meta-Analysis Jaime A. Davidson, MD1; Andreas Liebl, MD2; Jens S. Christiansen, MD3; Greg Fulcher, MD4; Robert J. Ligthelm, MD5; Paul Brown, CStat6; Titus Gylvin, MD7; and Ryuzo Kawamori, MD8 1Department
of Medicine, Division of Endocrinology, University of Texas, Southwestern Medical School, Dallas, Texas; 2Center for Diabetes and Metabolism, Fachklinik, Bad Heilbrunn, Germany; 3Medical Department M, Endocrinology and Diabetes, Aarhus University Hospital, Aarhus, Denmark; 4Department of Endocrinology, Diabetes and Metabolism, Royal North Shore Hospital, Sydney, Australia; 5EHM Clinic, Hoofddorp, The Netherlands; 6Novo Nordisk A/S, Soeborg, Denmark; 7Novo Nordisk Inc., Princeton, New Jersey; and 8Department of Medicine, Juntendo University, Tokyo, Japan ABSTRACT Background: Insulin is recommended as a secondline treatment after diet and metformin fail to reach and/or maintain glycemic targets considered to minimize the risk for long-term diabetic complications. Hypoglycemia and the fear of developing hypoglycemia, however, remain substantial barriers to the initiation and optimal use of insulin. Objective: The aim of this study was to compare biphasic insulin aspart 30 (BIAsp 30) with biphasic human insulin 30 (BHI 30) with respect to glycemic control and the risk for hypoglycemia using a metaanalysis of clinical trials comparing these insulins in patients with type 2 diabetes mellitus (T2DM). Methods: We included all published and unpublished, randomized, controlled trials in adult patients with T2DM (treatment duration ≥12 weeks) for which individual patient data were available. All clinical databases and local trial registries of Novo Nordisk A/S (Soeborg, Denmark) were searched to identify clinical trials comparing the 2 products. The predefined primary end point of the study was the overall rate of nocturnal hypoglycemia (major, minor, and symptoms-only hypoglycemia occurring from 12:00–6:00 am). Hypoglycemia was analyzed using a negative binomial distribution model, accounting for exposure time. Glycemic end points were analyzed at 12 to 16 weeks of treatment using ANCOVA, adjusting for baseline. Secondary safety end points were the rates of major hypoglycemia (hypoglycemia requiring third-party assistance), minor hypoglycemia (symptoms confirmed by plasma glucose [PG] <3.1 mmol/L), August 2009
daytime hypoglycemia (major, minor, and symptomsonly hypoglycemia occurring from 6:01 am–11:59 pm), overall hypoglycemia (the sum of all major, minor, and symptoms-only episodes), and change in weight from baseline to 12 to 16 weeks of treatment. Secondary efficacy end points were changes in glycosylated hemoglobin (HbA1c), fasting PG (FPG), postprandial PG increment (averaged over breakfast, lunch, and dinner), and insulin dose. Results: Nine randomized, parallel or crossover trials were included (N = 1674; male sex, 57%; mean [SD] age, 61.0 [10.6] years; body mass index, 26.7 [4.6] kg/m2; HbA1c, 8.1% [1.4%]; duration of diabetes, 10.9 [7.9] years). Rates of overall hypoglycemia were not significantly different (rate ratio [RR] = 1.08; 95% CI, 0.94–1.24; P = NS) between treatments. BIAsp 30 had a 50% lower rate of nocturnal hypoglycemia than BHI 30 (RR = 0.50; 95% CI, 0.38–0.67; P < 0.01), whereas the rate of daytime hypoglycemia was 24% lower for BHI 30 (RR = 1.24; 95% CI, 1.08–1.43; P < 0.01). The likelihood of major hypoglycemia was significantly lower with BIAsp 30 comData contained in this manuscript have been previously presented at the 43rd Annual Meeting of the European Association for the Study of Diabetes, September 17–21, 2007, Amsterdam, The Netherlands, and the 68th Annual American Diabetes Association Scientific Sessions, June 6–10, 2008, San Francisco, California. Accepted for publication May 14, 2009. doi:10.1016/j.clinthera.2009.08.011 0149-2918/$ - see front matter © 2009 Excerpta Medica Inc. All rights reserved.
1641
Clinical Therapeutics pared with BHI 30 (odds ratio = 0.45; 95% CI, 0.22–0.93; P < 0.05). BIAsp 30 was associated with reduced PPG increment (averaged over breakfast, lunch and dinner) compared with BHI 30 (treatment difference, –0.31; 95% CI, –0.49 to –0.07; P < 0.01). There was a significantly larger reduction in FPG associated with BHI 30 (treatment difference, 0.63; 95% CI, 0.31– 0.95; P < 0.01). However, no significant treatment difference was found for HbA1c (treatment difference, 0.04; 95% CI, –0.02 to 0.10; P = NS). Conclusion: This meta-analysis found BIAsp 30 to be associated with a significantly lower rate of nocturnal and major hypoglycemia, but a significantly increased risk for daytime hypoglycemia, compared with BHI 30 at a similar level of HbA1c in patients with T2DM. (Clin Ther. 2009;31:1641–1651) © 2009 Excerpta Medica Inc. Key words: biphasic insulin, meta-analysis, hypoglycemia, type 2 diabetes mellitus.
INTRODUCTION Type 2 diabetes mellitus (T2DM) is a progressive disease characterized by insulin resistance and a gradual decline in β-cell function that eventually necessitates the use of exogenous insulin by most patients. Although traditionally used as a final treatment option, insulin has been recently recommended by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) as a second-line treatment after diet and metformin fail to reach and/or maintain glycemic targets considered to minimize the risk for long-term diabetic complications.1 To attain these targets, an earlier introduction of insulin treatment has also been advocated by the American College of Endocrinology and the American Association of Clinical Endocrinologists.2 Hypoglycemia and the fear of developing hypoglycemia, however, remain substantial barriers to the initiation and optimal use of insulin.3 For some patients, another potential obstacle is the number of daily injections required by basal-bolus regimens. Because of this, premixed insulins containing varying proportions of rapid- and intermediate-acting insulin were developed to limit the number of daily injections for patients who require both prandial and basal insulin.4 The earliest insulin premixes contained neutral protein Hagedorn (NPH) insulin as the basal component and regular human insulin as the prandial com1642
ponent. More recently, premixed insulin analogues have been developed that more accurately mimic the mealtime components of endogenous insulin secretion, providing significantly better postprandial plasma glucose (PPG) control than conventional premixed human insulins.4 However, the relative rates of hypoglycemia associated with conventional and insulin analogue premixes are less clear. Biphasic insulin aspart 30 (BIAsp 30) is a premixed insulin analogue comprising 30% soluble rapid-acting insulin aspart and 70% intermediate-acting protaminecrystallized insulin aspart.5,6 Although some studies have shown BIAsp 30 to be associated with a significantly reduced rate of major7 and nocturnal hypoglycemia8 compared with biphasic human insulin 30 (BHI 30) (comprising 30% regular human insulin and 70% NPH insulin), other studies have been inconclusive.9–12 A possible reason for these different findings is that, apart from 1 study,8 none were designed to have the statistical power to detect a difference in rates of hypoglycemia between BIAsp 30 and BHI 30. Therefore, we conducted a meta-analysis of all published and unpublished, randomized, controlled trials for which individual patient data were available to compare the rate of hypoglycemia associated with BIAsp 30 and BHI 30 in patients with T2DM. Because there is a correlation between the degree of glycemic control and the risk for hypoglycemia,13 treatments were also compared with respect to specific glycemic end points.
SUBJECTS AND METHODS A retrospective meta-analysis of individual patient data was performed based on the guidelines of Whitehead14 and The Cochrane Policy Manual.15
Study Selection Criteria and Search Strategy All clinical databases and local trial registries of Novo Nordisk A/S, the manufacturer of BIAsp 30 (NovoLog® Mix 70/30 in the US and NovoMix® 30 elsewhere) and BHI 30 (Novolin® 70/30 in the US and Mixtard® 30 elsewhere), were searched to identify clinical trials comparing these 2 products. Predefined selection criteria were that the trial must be randomized and controlled, have a treatment duration of ≥12 weeks, individual patient data must be available, and the trial must have been conducted in accordance with the Declaration of Helsinki and the Guideline for Good Clinical Practice.16 Both published and unpublished studies were included. Short-term clinical pharVolume 31 Number 8
J.A. Davidson et al. macology trials (<12 weeks)10,11,17 and trials with different dosing regimens for the 2 insulin products (ie, BID vs TID18) were excluded.
Safety End Points The predefined primary end point was the overall rate of nocturnal hypoglycemia (all major, minor, and symptoms-only hypoglycemia occurring from 12:00– 6:00 am). Secondary safety end points were the rates of major hypoglycemia (hypoglycemia requiring thirdparty assistance), minor hypoglycemia (symptoms confirmed by PG <3.1 mmol/L), daytime hypoglycemia (all major, minor and symptoms-only hypoglycemia occurring from 6:01 am–11:59 pm), overall hypoglycemia (the sum of all major, minor, and symptoms-only episodes), and change in weight from baseline to 12 to 16 weeks of treatment. This was the treatment duration chosen to compare treatments with respect to glycemic end points and weight. The trials in the meta-analysis differed in length from 12 to 48 weeks (Table I). The
reason a 12- to 16-week window was chosen for this comparison was that not all trials included in the metaanalysis collected weight, insulin dose, and glycemic data at week 12; three trials collected these data at week 13 (trial 03819) or week 16 (trials 1394 and 14668). For all other trials, week-12 data were collected and used for the meta-analysis of weight, insulin dose, and glycemic end points. Hypoglycemic episodes were included in the analyses if they had an onset on or between the date of the first and the last dose of trial product. All analyses of hypoglycemia were based on self-reported episodes (symptoms-only or confirmed using standard blood glucose meters). For this meta-analysis, we decided to focus on confirmed hypoglycemia. Therefore, symptoms-only hypoglycemia was not a predefined end point of the study. It was not analyzed separately but was included in the meta-analysis of overall hypoglycemia (a combination of major, minor, and symptoms-only hypoglycemia).
Table I. Summary of randomized, controlled clinical trials comparing biphasic insulin aspart 30 (BIAsp 30) with biphasic human insulin 30 (BHI 30) included in the meta-analysis.* Treatment, Mean (SD), Daily Dose, U/kg Trial ID
Trial Design
Duration, Wk†
BIAsp 30
BHI 30
03819 108812 1234‡ 13539 1394§ 14668 1536§ 3002‡ 3006‡
Open, parallel Double-blind, parallel Double-blind, crossover Open, parallel Open, parallel Double-blind, crossover Open, parallel Double-blind, parallel Double-blind, parallel
13 14 12 48 16 16 12 12 12
0.56 (0.24) 0.15 (0.04) 0.75 (0.28) 0.41 (0.19) 0.64 (0.31) 0.66 (0.28) 0.54 (0.15) 0.80 (0.21) 0.69 (0.29)
0.58 (0.21) 0.14 (0.03) 0.74 (0.30) 0.40 (0.18) 0.64 (0.33) 0.64 (0.30) 0.56 (0.20) 0.81 (0.22) 0.78 (0.31)
*All trials were multicenter (6–55 centers), with 4 trials conducted in >1 country; all trials compared a twice-daily dosing regimen (before breakfast and dinner) of BIAsp 30 with BHI 30 except for trial 108812 that compared a once-daily (before dinner) regimen; none of the trials were treat-to-target (ie, insulin doses were not titrated according to any treatment algorithm); various inclusion criteria, such as certain glycosylated hemoglobin levels, age, and body mass index ranges and prestudy treatment requirements, were specified by the individual trials; with the exception of trial 1088,12 all patients were treated with insulin prior to trial entry. † Treatment duration for each period for crossover studies. ‡ Unpublished. § Unpublished, clinical trial synopsis available at www.novonordisk-trials.com.
August 2009
1643
Clinical Therapeutics
Efficacy End Points Secondary efficacy end points were changes from baseline to weeks 12 to 16 of treatment in glycosylated hemoglobin (HbA1c), fasting PG (FPG), PPG increment (averaged over breakfast, lunch, and dinner), and insulin dose. Data for FPG and PPG increment were derived from self-measured 8-point blood glucose profiles. PPG increment was defined as the PG concentration measured 90 minutes after a meal (120 minutes after a meal for trials 3002 and 3006) minus the PG concentration immediately prior to a meal.
ence (BIAsp 30 – BHI 30) with corresponding 95% CI was calculated for each end point. Heterogeneity was assessed for each end point by calculating I 2, the percentage of total variation across the studies that is due to heterogeneity rather than chance.21 Values >75% were considered to indicate high heterogeneity. If heterogeneity was observed for a particular end point, data were analyzed by a randomeffects meta-analysis model in addition to the fixedeffects model. An additional sensitivity analysis was conducted to explore the influence of blinding on conclusions for the primary end point, nocturnal hypoglycemia.
Statistical Methods The intent-to-treat population, defined as all randomized patients exposed to ≥1 dose of trial treatment,20 was used for the meta-analysis of all end points. For assessments that measured a change from baseline (weight and all efficacy end points), the last available measurement between 12 and 16 weeks was used. Missing values were imputed by the last-observationcarried-forward method. To combine results from crossover and parallel-group trials in the meta-analysis, estimates of the treatment effect were obtained separately for each study and then combined using the inverse variance method,14 whereby weight is assigned to the individual estimates according to their precision. The predefined primary objective of the meta-analysis was to compare BIAsp 30 and BHI 30 with respect to the rate of nocturnal hypoglycemia. Hypoglycemia was analyzed using a negative binomial distribution model, with the number of episodes as the dependent variable, log-transformed exposure time as an offset variable, baseline HbA1c as a continuous covariate, and treatment as fixed effect. The rate ratio (RR) of nocturnal hypoglycemia (BIAsp 30/BHI 30) was calculated together with 95% CI. Minor, daytime, and overall hypoglycemia was analyzed as described for nocturnal hypoglycemia. Due to the low number of events, the frequency of patients experiencing ≥1 episode of major hypoglycemia was compared between treatments using the Mantel-Haenszel method.14 An odds ratio (BIAsp 30/BHI 30) together with 95% CI was calculated. All outcomes that assessed a change from baseline to weeks 12 to 16 of treatment (weight, insulin dose, and all glycemic end points) were analyzed using ANCOVA, adjusting for appropriate baseline covariates (for crossover studies, patients were included as random effects). An estimate of the treatment differ1644
RESULTS Nine randomized, controlled trials were identified that met the selection criteria for the meta-analysis (Table I). The meta-analysis population comprised 1674 patients with T2DM (48.3% white, 49.6% Asian/ Pacific islander; male sex, 57.4%; mean [SD] age, 61.0 [10.6] years; body mass index, 26.7 [4.6] kg/m2; HbA1c, 8.1% [1.4%]; duration of diabetes, 10.9 [7.9] years) of which 810 were exposed to BIAsp 30, 524 to BHI 30, and 340 to both treatments (in crossover studies). The 2 treatment groups were closely matched with respect to demographic and baseline characteristics. Demographic characteristics are summarized on an individual trial basis in Table II.
Hypoglycemia Hypoglycemic episodes were reported in all 9 trials. The definition of major hypoglycemia was consistent across trials. For one trial (trial 13539), minor hypoglycemia was defined (and included in the meta-analysis) based on symptoms without a confirmatory measurement of PG concentration. The frequency of hypoglycemia in the meta-analysis population is summarized in Table III. No significant difference was found between treatments with respect to the rate of overall hypoglycemia (RR = 1.08; 95% CI, 0.94–1.24; P = NS; Table IV). However, BIAsp 30 had a significantly lower rate of nocturnal hypoglycemia than did BHI 30 (RR = 0.50; 95% CI, 0.38–0.67; P < 0.01; Figure 1), whereas BHI 30 was associated with a significantly lower rate of daytime hypoglycemia (RR = 1.24; 95% CI, 1.08–1.43; P < 0.01; Table IV). Significantly fewer patients experienced a major hypoglycemic episode with BIAsp 30 compared with BHI 30 (P < 0.05; Figure 2), while rates of minor hypoglycemia were not significantly different between treatments (Table IV). Volume 31 Number 8
August 2009 Table II. Demographic information for the randomized, controlled clinical trials included in the meta-analysis. Trial ID 03819 (n = 187)
108812 (n = 93)
1234 (n = 180)
13539 (n = 428)
1394 (n = 292)
14668 (n = 160)
95 (50.8) 92 (49.2)
39 (41.9) 54 (58.1)
79 (43.9) 101 (56.1)
177 (41.4) 251 (58.6)
125 (42.8) 167 (57.2)
48 (30.0) 112 (70.0)
185 (98.9) 1 (0.5) 1 (0.5) – –
68 (73.1) 10 (10.8) 1 (1.1) 11 (11.8) 3 (3.2)
112 (62.2) – 64 (35.6) – 4 (2.2)
– – 428 (100) – –
290 (99.3) 1 (0.3) 1 (0.3) – –
154 (96.3) 5 (3.1) 1 (0.6) – –
Weight, mean (SD), kg 79.2 (13.0)
85.4 (13.3)
77.7 (15.7)
60.0 (9.8)
Age, mean (SD), y
56.9 (11.5) 60.3 (12.0)
Sex, no. (%) Female Male Race, no. (%) White Black Asian/PI Hispanic Other
63.2 (8.6)
83.6 (14.6) 86.2 (14.5)
60.9 (10.0) 65.8 (10.2)
62.9 (9.1)
1536 (n = 195) 98 (50.3) 97 (49.7) – – 195 (100) – –
3002 (n = 36) 16 (44.4) 20 (55.6) – – 36 (100) – –
3006 (n = 103)
46 (44.7) 713 (42.6) 57 (55.3) 961 (57.4) – – 103 (100) – –
66.6 (10.0) 63.7 (10.0) 63.5 (10.0) 57.8 (8.7)
Total* (N = 1674)
809 (48.3) 17 (1.0) 830 (49.6) 11 (0.7) 7 (0.4) –
53.5 (10.7) 53.8 (10.6) 61.0 (10.6)
PI = Pacific islander. *Patients with type 2 diabetes mellitus randomized and exposed to treatment.
J.A. Davidson et al.
1645
Clinical Therapeutics
Table III. Summary of hypoglycemic episodes in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30) (N = 1674). BIAsp 30 (n = 1150)* Episode Overall hypoglycemia† Major hypoglycemia Minor hypoglycemia Daytime hypoglycemia‡§ Nocturnal hypoglycemia§=
BHI 30 (n = 864)*
No. (%)
Episodes
No. (%)
Episodes
551 (47.9) 13 (1.1) 350 (30.4) 533 (46.3) 127 (11.0)
3205 18 2047 2782 369
382 (44.2) 20 (2.3) 203 (23.5) 343 (39.7) 147 (17.0)
2234 29 1234 1807 404
*Patients exposed to study drug (including 340 who received both in crossover studies). † Major, minor, and symptoms-only hypoglycemia. ‡ 6:01 AM to 11:59 PM. § Due to missing information, some hypoglycemic episodes (~2.5%) could not be assigned to either daytime or nocturnal. =12:00 AM to 6:00 AM.
Table IV. Meta-analyses of secondary end points in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30); rate ratios and treatment differences were calculated by the fixed-effects model. Variable
Rate Ratio* or Treatment Difference†
P
Heterogeneity (I2), %
Overall hypoglycemia‡ Minor hypoglycemia Daytime hypoglycemia§ HbA1c FPG PPG increment = Weight Insulin dose¶
1.08 (0.94 to 1.24)* 0.98 (0.81 to 1.20)* 1.24 (1.08 to 1.43)* 0.04 (–0.02 to 0.10)† 0.63 (0.31 to 0.95)† –0.31 (–0.49 to –0.07)† –0.07 (–0.24 to 0.09)† 0.01 (0.01 to 0.02)†
0.27 0.87 <0.01 0.25 <0.01 <0.01 0.40 <0.01
23 0 49 0 0 0 58 35
HbA1c = glycosylated hemoglobin; FPG = fasting plasma glucose; PPG = postprandial plasma glucose. * BIAsp 30/BHI 30. † BIAsp 30 – BHI 30. ‡ Major, minor, and symptoms-only hypoglycemia. § 6:01 AM to 11:59 PM. =Averaged over breakfast, lunch, and dinner. ¶ Only data for the first treatment period were included for the 2 crossover trials.
A sensitivity analysis, in which open-label studies were excluded from the meta-analysis of nocturnal hypoglycemia, yielded a similar result (RR = 0.41; 95% CI, 0.29–0.59; P < 0.01; I 2 = 34%) to that obtained when all trials were included (Figure 1). 1646
Glycemic Control All 9 trials provided HbA1c data for the metaanalysis. Three trials (1353,9 1466,8 and 1536) were excluded from the meta-analyses of PPG increment and FPG, as these data were not collected. Mean baseVolume 31 Number 8
J.A. Davidson et al.
Trial
Rate Ratio (95% CI)
P
03819
0.57 (0.20–1.58)
0.28
108812
0.89 (0.25–3.16)
0.86
1234
0.44 (0.22–0.89)
0.02
13539
1.03 (0.42–2.53)
0.95
1394
1.03 (0.38–2.76)
0.96
14668
0.33 (0.21–0.51)
0.01
1536
0.44 (0.11–1.47)
0.17
3002
1.05 (0.11–10.09)
0.97
3006
2.43 (0.31–18.90)
0.39
Overall*
0.50 (0.38–0.67)
<0.01
Test of Heterogeneity: I2 = 32% 0.1 0.2
1
Favors BIAsp 30
10
20
Favors BHI 30
Figure 1. Meta-analysis of nocturnal hypoglycemia in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30). *Data for all trials combined (fixed-effects estimate).
line and end-of-treatment values for HbA1c, FPG, and PPG increment are presented in Table V. BIAsp 30 was associated with reduced PPG increment (averaged over breakfast, lunch, and dinner) by a significantly greater extent than was BHI 30 (P < 0.01; Table IV), whereas BHI 30 treatment was associated with a significantly larger reduction in FPG (P < 0.01). However, no significant treatment difference was found for HbA1c.
Weight Eight clinical trials provided weight data for the meta-analysis. One trial (14668) was excluded because data were only collected at baseline. For trial 1234, change in weight from baseline was determined separately for each period. Both BIAsp 30 and BHI 30 were associated with an increase in weight from baseAugust 2009
line (mean [SD] change of 0.2 [2.3] and 0.7 [2.1] kg, respectively), but no significant treatment difference was found (Table IV).
Insulin Dose All 9 trials provided total daily insulin dose data for the meta-analysis. A higher increase from baseline in total daily dose was observed for BIAsp 30 relative to BHI 30 (P < 0.01; Table IV). For BIAsp 30, total daily dose increased from 0.54 (0.28) U/kg at baseline to 0.61 (0.31) U/kg; for BHI 30, total daily dose increased from 0.58 (0.30) U/kg at baseline to 0.64 (0.32) U/kg.
Heterogeneity In all cases where heterogeneity was found across trials for a particular end point (ie, I 2 >0%), similar 1647
Clinical Therapeutics
Trial
Odds Ratio (95% CI)
P
03819
0.50 (0.12–1.98)
0.32
1234
0.34 (0.01–8.28)
0.50
13539
0.57 (0.16–2.00)
0.38
1394
0.53 (0.03–8.63)
0.66
14668
0.31 (0.06–1.54)
0.15
3002
0.25 (0.01–6.62)
0.41
Overall*
0.45 (0.22–0.93)
<0.05
Test of Heterogeneity: I2 = 0% 0.01
0.1
1
Favors BIAsp 30
10
100
Favors BHI 30
Figure 2. Meta-analysis of major hypoglycemia in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30). *Data for all trials combined (fixed-effects estimate).
overall estimates and 95% CI were obtained when the meta-analysis was repeated using a random-effects model (data not shown).
DISCUSSION Intensive insulin treatment, designed to reduce the risk of long-term diabetic complications, has been associated with an increased risk of hypoglycemia in patients with T2DM, with clinical consequences ranging from mild discomfort to coma and even death.22 Although considered less common in patients with T2DM, rates of hypoglycemia may approach those observed in patients with type 1 diabetes with increasing age, duration of diabetes, and length of insulin treatment.23 Fear of hypoglycemia can result in a reluctance to initiate insulin treatment, even though insulin is recog1648
nized as the most effective diabetes medication to lower glycemia.1 Moreover, cases have been reported where patients intentionally keep blood glucose levels above recommended targets to avoid hypoglycemia.3 It is important, therefore, to evaluate insulin products and/or treatment regimens that carry a lower risk of hypoglycemia and do not jeopardize glycemic efficacy. In this study, we compared the rate of hypoglycemia associated with BIAsp 30 and BHI 30 using a meta-analysis of 9 published and unpublished randomized, controlled trials in adults with T2DM. Four of these trials also included patients with type 1 diabetes, but there was an insufficient number (151 patients) to conduct a meaningful meta-analysis on this patient population. Our results, based on individual data for 1674 patients with T2DM, suggested that, Volume 31 Number 8
J.A. Davidson et al.
Table V. Summary of glycemic end points in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30). BIAsp 30
HbA1c, % Baseline Weeks 12 to 16 Change from baseline to weeks 12 to 16 FPG, mmol/L Baseline Weeks 12 to 16 Change from baseline to weeks 12 to 16 PPG increment, mmol/L* Baseline Weeks 12 to 16 Change from baseline to weeks 12 to 16
BHI 30
No. of Patients
Mean (SD)
No. of Patients
Mean (SD)
965 929
8.1 (1.4) 7.9 (1.2)
690 647
8.3 (1.4) 7.9 (1.2)
929
–0.2 (1.0)
647
–0.3 (1.0)
466 458
9.4 (3.2) 9.4 (2.7)
400 396
9.6 (3.7) 8.9 (3.5)
447
0 (3.6)
389
–0.8 (4.2)
446 446
2.98 (2.22) 2.19 (2.26)
386 391
3.27 (2.47) 2.71 (2.24)
421
–0.81 (2.43)
372
–0.53 (2.58)
HbA1c = glycosylated hemoglobin; FPG = fasting plasma glucose; PPG = postprandial plasma glucose. *Averaged over breakfast, lunch, and dinner.
although no statistically significant difference in the overall rate of hypoglycemia was found between treatments, BIAsp 30 was associated with a significantly lower rate of nocturnal hypoglycemia compared with BHI 30, but a significantly higher rate of daytime hypoglycemia. Hypoglycemia is unwelcome at any time of the day but may be a particular problem at night, because a sleeping patient may be unaware of symptoms that would have otherwise prompted corrective action. Whereas no significant difference was found in the frequency of minor hypoglycemia associated with BIAsp 30 or BHI 30, significantly fewer patients on BIAsp 30 experienced a major hypoglycemic event. Because major hypoglycemia can be potentially life threatening (and incur high health care costs), reducing the risk for such events is essential for both the patient and health care industry.24 In terms of glycemic control, the meta-analysis found that BIAsp 30 was associated with reduced PPG increment averaged over the 3 main meals to a signifiAugust 2009
cantly greater extent than was BHI 30, whereas a significantly larger reduction in FPG was found for BHI 30. However, overall metabolic control (as measured by HbA1c) did not differ significantly between treatment groups. Therefore, the reduced risk for nocturnal and major hypoglycemia associated with BIAsp 30 was not achieved at the expense of overall glycemic control. It is also noteworthy that the comparable overall glycemic control provided by BIAsp 30 and BHI 30 was obtained with similar total daily doses of insulin, and that both treatments resulted in similar weight gain. For both treatment groups, HbA1c following 12 to 16 weeks of treatment was ~7.9%. This is above ADA- and EASD-recommended HbA1c targets of <7%25 and may reflect the fact that none of the trials in the meta-analysis were of a treat-to-target design. Due to the relatively short treatment duration of the majority of clinical trials in the meta-analysis (12 to 16 weeks) and the low number of major hypoglycemic episodes reported in these studies, our findings for HbA1c and major hypoglycemia need to be 1649
Clinical Therapeutics confirmed by a larger, properly powered, randomized, controlled trial of sufficient length. The meta-analysis confirmed previous findings that BIAsp 30 provides significantly better postprandial glycemic control than does BHI 30.9–11,17 The improved mealtime coverage has been attributed to the faster onset and higher plasma insulin levels attained by the soluble insulin aspart component of BIAsp 30 compared with the regular human insulin constituent of BHI 30.4 Shortacting human insulin has a longer pharmacodynamic action profile than does insulin aspart,26 with this characteristic likely to account, at least in part, for the significantly higher rate of nocturnal hypoglycemia with BHI 30 in this study. The observed peaking action of the intermediate-acting NPH insulin component of BHI 30 may also contribute to the higher rate of nocturnal hypoglycemia and lower FPG levels observed for this premixed insulin.27 As higher prandial plasma insulin levels are attained with BIAsp 30 than BHI 30,4 it could be speculated that the higher rate of daytime hypoglycemia associated with this insulin premix may be partially attributed to occasions in which there is inadequate carbohydrate intake with the mealtime dose. Low to moderate heterogeneity was found across trials for some of the end points analyzed (weight, insulin dose, overall hypoglycemia, daytime hypoglycemia, and nocturnal hypoglycemia). However, in all cases, the same conclusions were reached for the end point regardless of whether a fixed- or random-effects model was used for the meta-analysis, thereby supporting the robustness of the results. To test for any potential bias in the reporting of hypoglycemic events for open-label trials, a sensitivity analysis was performed for the primary end point whereby only trials that were blinded to treatment allocation were included in the meta-analysis of nocturnal hypoglycemia. Near-identical results were found to those observed for the full meta-analysis dataset, indicating that reporting bias was unlikely.
CONCLUSION This meta-analysis found BIAsp 30 to be associated with a significantly lower rate of nocturnal and major hypoglycemia, but a significantly increased risk for daytime hypoglycemia, compared with BHI 30 at a similar level of HbA1c in patients with T2DM.
ACKNOWLEDGMENTS The authors wish to thank the study coordinators, nurses, physicians, and patients who participated in 1650
the clinical trials included in this meta-analysis. We also thank Paul G. Drake, PhD, Novo Nordisk A/S, who assisted in the preparation of this manuscript. Dr. Davidson has been a consultant (speaker, advisory board member, clinical trials) for Novo Nordisk. Dr. Liebl has received honoraria from Novo Nordisk for scientific presentations and participating at advisory board meetings, and has received financial help with clinical studies from Novo Nordisk. Dr. Christiansen has received research grants and speakers’ fees from Novo Nordisk. Dr. Fulcher has received honoraria from Novo Nordisk for speaking at educational activities and participating at advisory board meetings. Dr. Ligthelm has been a member of Novo Nordisk advisory boards and has been a speaker at Novo Nordisk meetings. Dr. Gylvin owns shares in Novo Nordisk. Dr. Kawamori has received honoraria for lectures cosponsored by Novo Nordisk.
REFERENCES 1. Nathan DM, Buse JB, Davidson MB, et al. Management of hyperglycemia in type 2 diabetes: A consensus algorithm for the initiation and adjustment of therapy: A consensus statement from the American Diabetes Association and the European Association for the Study of Diabetes [published correction appears in Diabetes Care. 2006;49:2816–2818]. Diabetes Care. 2006;29:1963–1972. 2. Jellinger PS, Davidson JA, Blonde L, et al, for the ACE/ AACE Diabetes Road Map Task Force. Road maps to achieve glycemic control in type 2 diabetes mellitus: ACE/ AACE Diabetes Road Map Task Force. Endocr Pract. 2007;13:260–268. 3. Wild D, von Maltzahn R, Brohan E, et al. A critical review of the literature on fear of hypoglycemia in diabetes: Implications for diabetes management and patient education. Patient Educ Couns. 2007;68:10–15. 4. Garber AJ, Ligthelm R, Christiansen JS, Liebl A. Premixed insulin treatment for type 2 diabetes: Analogue or human? Diabetes Obes Metab. 2007;9:630–639. 5. Halimi S, Raskin P, Liebl A, et al. Efficacy of biphasic insulin aspart in patients with type 2 diabetes. Clin Ther. 2005;27(Suppl B):S57–S74. 6. Gough SC, Tibaldi J. Biphasic insulin aspart in type 2 diabetes mellitus: An evidence-based medicine review. Clin Drug Investig. 2007;27:299–324. 7. Boehm BO, Vaz JA, Brøndsted L, Home PD. Long-term efficacy and safety of biphasic insulin aspart in patients with type 2 diabetes. Eur J Intern Med. 2004;15:496–502. 8. McNally PG, Dean JD, Morris AD, et al. Using continuous glucose monitoring to measure the frequency of low glucose values when using biphasic insulin aspart 30 com-
Volume 31 Number 8
J.A. Davidson et al.
9.
10.
11.
12.
13.
14.
15.
16.
pared with biphasic human insulin 30: A double-blind crossover study in individuals with type 2 diabetes. Diabetes Care. 2007;30:1044–1048. Iwamoto Y. A randomised, multicentre trial of biphasic insulin aspart versus biphasic human insulin in type 2 diabetes. Diabetologia. 2003; 46:A270. McSorley PT, Bell PM, Jacobsen LV, et al. Twice-daily biphasic insulin aspart 30 versus biphasic human insulin 30: A double-blind crossover study in adults with type 2 diabetes mellitus. Clin Ther. 2002;24:530–539. Hermansen K, Colombo M, Storgaard H, et al. Improved postprandial glycemic control with biphasic insulin aspart relative to biphasic insulin lispro and biphasic human insulin in patients with type 2 diabetes. Diabetes Care. 2002;25:883– 888. Kilo C, Mezitis N, Jain R, et al. Starting patients with type 2 diabetes on insulin therapy using oncedaily injections of biphasic insulin aspart 70/30, biphasic human insulin 70/30, or NPH insulin in combination with metformin. J Diabetes Complications. 2003;17:307–313. Wright AD, Cull CA, Macleod KM, et al. Hypoglycemia in type 2 diabetic patients randomized to and maintained on monotherapy with diet, sulfonylurea, metformin, or insulin for 6 years from diagnosis: UKPDS73. J Diabetes Complications. 2006;20:395–401. Whitehead A, ed. Meta-Analysis of Controlled Clinical Trials. Chichester, UK: John Wiley & Sons; 2002. The Cochrane Collaboration. The Cochrane Policy Manual. Issue 3. http://www.cochrane.org/admin/ manual.htm. Accessed July 16, 2009. ICH Harmonised Tripartite Guideline: Guideline for Good Clinical Practice E6(R1). June 1996. http:// www.emea.europa.eu/pdf s/ human/ich/013595en.pdf. Accessed January 16, 2009.
August 2009
17. Kapitza C, Rave K, Ostrowski K, et al. Reduced postprandial glycaemic excursion with biphasic insulin Aspart 30 injected immediately before a meal. Diabet Med. 2004;21:500– 501. 18. Abrahamian H, Ludvik B, Schernthaner G, et al. Improvement of glucose tolerance in type 2 diabetic patients: Traditional vs modern insulin regimens (results from the Austrian Biaspart Study). Horm Metab Res. 2005;37:684–689. 19. Boehm BO, Home PD, Behrend C, et al. Premixed insulin aspart 30 vs premixed human insulin 30/70 twice daily: A randomized trial in type 1 and type 2 diabetic patients [published correction appears in Diabet Med. 2002;19:797]. Diabet Med. 2002; 19:393–399. 20. International Conference on Harmonisation (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Harmonised Tripartite Guideline. Statistical principles for clinical trials: E9. http://www.ich.org/LOB/ media/MEDIA485.pdf. Accessed July 9, 2009. 21. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327: 557–560. 22. Cryer PE, Davis SN, Shamoon H. Hypoglycemia in diabetes. Diabetes Care. 2003;26:1902–1912.
23. UK Hypoglycaemia Study Group, ROYAUME-UNI. Risk of hypoglycaemia in types 1 and 2 diabetes: Effects of treatment modalities and their duration. Diabetologia. 2007;50: 1140–1147. 24. Amiel SA, Dixon T, Mann R, Jameson K. Hypoglycaemia in type 2 diabetes. Diabet Med. 2008;25:245–254. 25. Nathan DM, Buse JB, Davidson MB, et al, for the European Association for the Study of Diabetes. 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. 2009;32:193–203. 26. Mudaliar SR, Lindberg FA, Joyce M, et al. Insulin aspart (B28 asp-insulin): A fast-acting analog of human insulin: Absorption kinetics and action profile compared with regular human insulin in healthy nondiabetic subjects. Diabetes Care. 1999;22: 1501–1506. 27. Heise T, Heinemann L, Hövelmann U, et al. Biphasic insulin aspart 30/70 (BIAsp 30): Pharmacokinetics (PK) and pharmacodynamics (PD) in comparison with once-daily biphasic human insulin and basalbolus therapy [Published online ahead of print June 1, 2009]. Diabetes Care.
Address correspondence to: Paul Brown, CStat, Department of Biostatistics, Novo Nordisk A/S, Vantaarnsvej 114, 2860 Soeborg, Denmark. E-mail: [email protected] 1651
Risk for Nocturnal Hypoglycemia With Biphasic Insulin Aspart 30 Compared With Biphasic Human Insulin 30 in Adults With Type 2 Diabetes Mellitus: A Meta-Analysis Jaime A. Davidson, MD1; Andreas Liebl, MD2; Jens S. Christiansen, MD3; Greg Fulcher, MD4; Robert J. Ligthelm, MD5; Paul Brown, CStat6; Titus Gylvin, MD7; and Ryuzo Kawamori, MD8 1Department
of Medicine, Division of Endocrinology, University of Texas, Southwestern Medical School, Dallas, Texas; 2Center for Diabetes and Metabolism, Fachklinik, Bad Heilbrunn, Germany; 3Medical Department M, Endocrinology and Diabetes, Aarhus University Hospital, Aarhus, Denmark; 4Department of Endocrinology, Diabetes and Metabolism, Royal North Shore Hospital, Sydney, Australia; 5EHM Clinic, Hoofddorp, The Netherlands; 6Novo Nordisk A/S, Soeborg, Denmark; 7Novo Nordisk Inc., Princeton, New Jersey; and 8Department of Medicine, Juntendo University, Tokyo, Japan ABSTRACT Background: Insulin is recommended as a secondline treatment after diet and metformin fail to reach and/or maintain glycemic targets considered to minimize the risk for long-term diabetic complications. Hypoglycemia and the fear of developing hypoglycemia, however, remain substantial barriers to the initiation and optimal use of insulin. Objective: The aim of this study was to compare biphasic insulin aspart 30 (BIAsp 30) with biphasic human insulin 30 (BHI 30) with respect to glycemic control and the risk for hypoglycemia using a metaanalysis of clinical trials comparing these insulins in patients with type 2 diabetes mellitus (T2DM). Methods: We included all published and unpublished, randomized, controlled trials in adult patients with T2DM (treatment duration ≥12 weeks) for which individual patient data were available. All clinical databases and local trial registries of Novo Nordisk A/S (Soeborg, Denmark) were searched to identify clinical trials comparing the 2 products. The predefined primary end point of the study was the overall rate of nocturnal hypoglycemia (major, minor, and symptoms-only hypoglycemia occurring from 12:00–6:00 am). Hypoglycemia was analyzed using a negative binomial distribution model, accounting for exposure time. Glycemic end points were analyzed at 12 to 16 weeks of treatment using ANCOVA, adjusting for baseline. Secondary safety end points were the rates of major hypoglycemia (hypoglycemia requiring third-party assistance), minor hypoglycemia (symptoms confirmed by plasma glucose [PG] <3.1 mmol/L), August 2009
daytime hypoglycemia (major, minor, and symptomsonly hypoglycemia occurring from 6:01 am–11:59 pm), overall hypoglycemia (the sum of all major, minor, and symptoms-only episodes), and change in weight from baseline to 12 to 16 weeks of treatment. Secondary efficacy end points were changes in glycosylated hemoglobin (HbA1c), fasting PG (FPG), postprandial PG increment (averaged over breakfast, lunch, and dinner), and insulin dose. Results: Nine randomized, parallel or crossover trials were included (N = 1674; male sex, 57%; mean [SD] age, 61.0 [10.6] years; body mass index, 26.7 [4.6] kg/m2; HbA1c, 8.1% [1.4%]; duration of diabetes, 10.9 [7.9] years). Rates of overall hypoglycemia were not significantly different (rate ratio [RR] = 1.08; 95% CI, 0.94–1.24; P = NS) between treatments. BIAsp 30 had a 50% lower rate of nocturnal hypoglycemia than BHI 30 (RR = 0.50; 95% CI, 0.38–0.67; P < 0.01), whereas the rate of daytime hypoglycemia was 24% lower for BHI 30 (RR = 1.24; 95% CI, 1.08–1.43; P < 0.01). The likelihood of major hypoglycemia was significantly lower with BIAsp 30 comData contained in this manuscript have been previously presented at the 43rd Annual Meeting of the European Association for the Study of Diabetes, September 17–21, 2007, Amsterdam, The Netherlands, and the 68th Annual American Diabetes Association Scientific Sessions, June 6–10, 2008, San Francisco, California. Accepted for publication May 14, 2009. doi:10.1016/j.clinthera.2009.08.011 0149-2918/$ - see front matter © 2009 Excerpta Medica Inc. All rights reserved.
1641
Clinical Therapeutics pared with BHI 30 (odds ratio = 0.45; 95% CI, 0.22–0.93; P < 0.05). BIAsp 30 was associated with reduced PPG increment (averaged over breakfast, lunch and dinner) compared with BHI 30 (treatment difference, –0.31; 95% CI, –0.49 to –0.07; P < 0.01). There was a significantly larger reduction in FPG associated with BHI 30 (treatment difference, 0.63; 95% CI, 0.31– 0.95; P < 0.01). However, no significant treatment difference was found for HbA1c (treatment difference, 0.04; 95% CI, –0.02 to 0.10; P = NS). Conclusion: This meta-analysis found BIAsp 30 to be associated with a significantly lower rate of nocturnal and major hypoglycemia, but a significantly increased risk for daytime hypoglycemia, compared with BHI 30 at a similar level of HbA1c in patients with T2DM. (Clin Ther. 2009;31:1641–1651) © 2009 Excerpta Medica Inc. Key words: biphasic insulin, meta-analysis, hypoglycemia, type 2 diabetes mellitus.
INTRODUCTION Type 2 diabetes mellitus (T2DM) is a progressive disease characterized by insulin resistance and a gradual decline in β-cell function that eventually necessitates the use of exogenous insulin by most patients. Although traditionally used as a final treatment option, insulin has been recently recommended by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) as a second-line treatment after diet and metformin fail to reach and/or maintain glycemic targets considered to minimize the risk for long-term diabetic complications.1 To attain these targets, an earlier introduction of insulin treatment has also been advocated by the American College of Endocrinology and the American Association of Clinical Endocrinologists.2 Hypoglycemia and the fear of developing hypoglycemia, however, remain substantial barriers to the initiation and optimal use of insulin.3 For some patients, another potential obstacle is the number of daily injections required by basal-bolus regimens. Because of this, premixed insulins containing varying proportions of rapid- and intermediate-acting insulin were developed to limit the number of daily injections for patients who require both prandial and basal insulin.4 The earliest insulin premixes contained neutral protein Hagedorn (NPH) insulin as the basal component and regular human insulin as the prandial com1642
ponent. More recently, premixed insulin analogues have been developed that more accurately mimic the mealtime components of endogenous insulin secretion, providing significantly better postprandial plasma glucose (PPG) control than conventional premixed human insulins.4 However, the relative rates of hypoglycemia associated with conventional and insulin analogue premixes are less clear. Biphasic insulin aspart 30 (BIAsp 30) is a premixed insulin analogue comprising 30% soluble rapid-acting insulin aspart and 70% intermediate-acting protaminecrystallized insulin aspart.5,6 Although some studies have shown BIAsp 30 to be associated with a significantly reduced rate of major7 and nocturnal hypoglycemia8 compared with biphasic human insulin 30 (BHI 30) (comprising 30% regular human insulin and 70% NPH insulin), other studies have been inconclusive.9–12 A possible reason for these different findings is that, apart from 1 study,8 none were designed to have the statistical power to detect a difference in rates of hypoglycemia between BIAsp 30 and BHI 30. Therefore, we conducted a meta-analysis of all published and unpublished, randomized, controlled trials for which individual patient data were available to compare the rate of hypoglycemia associated with BIAsp 30 and BHI 30 in patients with T2DM. Because there is a correlation between the degree of glycemic control and the risk for hypoglycemia,13 treatments were also compared with respect to specific glycemic end points.
SUBJECTS AND METHODS A retrospective meta-analysis of individual patient data was performed based on the guidelines of Whitehead14 and The Cochrane Policy Manual.15
Study Selection Criteria and Search Strategy All clinical databases and local trial registries of Novo Nordisk A/S, the manufacturer of BIAsp 30 (NovoLog® Mix 70/30 in the US and NovoMix® 30 elsewhere) and BHI 30 (Novolin® 70/30 in the US and Mixtard® 30 elsewhere), were searched to identify clinical trials comparing these 2 products. Predefined selection criteria were that the trial must be randomized and controlled, have a treatment duration of ≥12 weeks, individual patient data must be available, and the trial must have been conducted in accordance with the Declaration of Helsinki and the Guideline for Good Clinical Practice.16 Both published and unpublished studies were included. Short-term clinical pharVolume 31 Number 8
J.A. Davidson et al. macology trials (<12 weeks)10,11,17 and trials with different dosing regimens for the 2 insulin products (ie, BID vs TID18) were excluded.
Safety End Points The predefined primary end point was the overall rate of nocturnal hypoglycemia (all major, minor, and symptoms-only hypoglycemia occurring from 12:00– 6:00 am). Secondary safety end points were the rates of major hypoglycemia (hypoglycemia requiring thirdparty assistance), minor hypoglycemia (symptoms confirmed by PG <3.1 mmol/L), daytime hypoglycemia (all major, minor and symptoms-only hypoglycemia occurring from 6:01 am–11:59 pm), overall hypoglycemia (the sum of all major, minor, and symptoms-only episodes), and change in weight from baseline to 12 to 16 weeks of treatment. This was the treatment duration chosen to compare treatments with respect to glycemic end points and weight. The trials in the meta-analysis differed in length from 12 to 48 weeks (Table I). The
reason a 12- to 16-week window was chosen for this comparison was that not all trials included in the metaanalysis collected weight, insulin dose, and glycemic data at week 12; three trials collected these data at week 13 (trial 03819) or week 16 (trials 1394 and 14668). For all other trials, week-12 data were collected and used for the meta-analysis of weight, insulin dose, and glycemic end points. Hypoglycemic episodes were included in the analyses if they had an onset on or between the date of the first and the last dose of trial product. All analyses of hypoglycemia were based on self-reported episodes (symptoms-only or confirmed using standard blood glucose meters). For this meta-analysis, we decided to focus on confirmed hypoglycemia. Therefore, symptoms-only hypoglycemia was not a predefined end point of the study. It was not analyzed separately but was included in the meta-analysis of overall hypoglycemia (a combination of major, minor, and symptoms-only hypoglycemia).
Table I. Summary of randomized, controlled clinical trials comparing biphasic insulin aspart 30 (BIAsp 30) with biphasic human insulin 30 (BHI 30) included in the meta-analysis.* Treatment, Mean (SD), Daily Dose, U/kg Trial ID
Trial Design
Duration, Wk†
BIAsp 30
BHI 30
03819 108812 1234‡ 13539 1394§ 14668 1536§ 3002‡ 3006‡
Open, parallel Double-blind, parallel Double-blind, crossover Open, parallel Open, parallel Double-blind, crossover Open, parallel Double-blind, parallel Double-blind, parallel
13 14 12 48 16 16 12 12 12
0.56 (0.24) 0.15 (0.04) 0.75 (0.28) 0.41 (0.19) 0.64 (0.31) 0.66 (0.28) 0.54 (0.15) 0.80 (0.21) 0.69 (0.29)
0.58 (0.21) 0.14 (0.03) 0.74 (0.30) 0.40 (0.18) 0.64 (0.33) 0.64 (0.30) 0.56 (0.20) 0.81 (0.22) 0.78 (0.31)
*All trials were multicenter (6–55 centers), with 4 trials conducted in >1 country; all trials compared a twice-daily dosing regimen (before breakfast and dinner) of BIAsp 30 with BHI 30 except for trial 108812 that compared a once-daily (before dinner) regimen; none of the trials were treat-to-target (ie, insulin doses were not titrated according to any treatment algorithm); various inclusion criteria, such as certain glycosylated hemoglobin levels, age, and body mass index ranges and prestudy treatment requirements, were specified by the individual trials; with the exception of trial 1088,12 all patients were treated with insulin prior to trial entry. † Treatment duration for each period for crossover studies. ‡ Unpublished. § Unpublished, clinical trial synopsis available at www.novonordisk-trials.com.
August 2009
1643
Clinical Therapeutics
Efficacy End Points Secondary efficacy end points were changes from baseline to weeks 12 to 16 of treatment in glycosylated hemoglobin (HbA1c), fasting PG (FPG), PPG increment (averaged over breakfast, lunch, and dinner), and insulin dose. Data for FPG and PPG increment were derived from self-measured 8-point blood glucose profiles. PPG increment was defined as the PG concentration measured 90 minutes after a meal (120 minutes after a meal for trials 3002 and 3006) minus the PG concentration immediately prior to a meal.
ence (BIAsp 30 – BHI 30) with corresponding 95% CI was calculated for each end point. Heterogeneity was assessed for each end point by calculating I 2, the percentage of total variation across the studies that is due to heterogeneity rather than chance.21 Values >75% were considered to indicate high heterogeneity. If heterogeneity was observed for a particular end point, data were analyzed by a randomeffects meta-analysis model in addition to the fixedeffects model. An additional sensitivity analysis was conducted to explore the influence of blinding on conclusions for the primary end point, nocturnal hypoglycemia.
Statistical Methods The intent-to-treat population, defined as all randomized patients exposed to ≥1 dose of trial treatment,20 was used for the meta-analysis of all end points. For assessments that measured a change from baseline (weight and all efficacy end points), the last available measurement between 12 and 16 weeks was used. Missing values were imputed by the last-observationcarried-forward method. To combine results from crossover and parallel-group trials in the meta-analysis, estimates of the treatment effect were obtained separately for each study and then combined using the inverse variance method,14 whereby weight is assigned to the individual estimates according to their precision. The predefined primary objective of the meta-analysis was to compare BIAsp 30 and BHI 30 with respect to the rate of nocturnal hypoglycemia. Hypoglycemia was analyzed using a negative binomial distribution model, with the number of episodes as the dependent variable, log-transformed exposure time as an offset variable, baseline HbA1c as a continuous covariate, and treatment as fixed effect. The rate ratio (RR) of nocturnal hypoglycemia (BIAsp 30/BHI 30) was calculated together with 95% CI. Minor, daytime, and overall hypoglycemia was analyzed as described for nocturnal hypoglycemia. Due to the low number of events, the frequency of patients experiencing ≥1 episode of major hypoglycemia was compared between treatments using the Mantel-Haenszel method.14 An odds ratio (BIAsp 30/BHI 30) together with 95% CI was calculated. All outcomes that assessed a change from baseline to weeks 12 to 16 of treatment (weight, insulin dose, and all glycemic end points) were analyzed using ANCOVA, adjusting for appropriate baseline covariates (for crossover studies, patients were included as random effects). An estimate of the treatment differ1644
RESULTS Nine randomized, controlled trials were identified that met the selection criteria for the meta-analysis (Table I). The meta-analysis population comprised 1674 patients with T2DM (48.3% white, 49.6% Asian/ Pacific islander; male sex, 57.4%; mean [SD] age, 61.0 [10.6] years; body mass index, 26.7 [4.6] kg/m2; HbA1c, 8.1% [1.4%]; duration of diabetes, 10.9 [7.9] years) of which 810 were exposed to BIAsp 30, 524 to BHI 30, and 340 to both treatments (in crossover studies). The 2 treatment groups were closely matched with respect to demographic and baseline characteristics. Demographic characteristics are summarized on an individual trial basis in Table II.
Hypoglycemia Hypoglycemic episodes were reported in all 9 trials. The definition of major hypoglycemia was consistent across trials. For one trial (trial 13539), minor hypoglycemia was defined (and included in the meta-analysis) based on symptoms without a confirmatory measurement of PG concentration. The frequency of hypoglycemia in the meta-analysis population is summarized in Table III. No significant difference was found between treatments with respect to the rate of overall hypoglycemia (RR = 1.08; 95% CI, 0.94–1.24; P = NS; Table IV). However, BIAsp 30 had a significantly lower rate of nocturnal hypoglycemia than did BHI 30 (RR = 0.50; 95% CI, 0.38–0.67; P < 0.01; Figure 1), whereas BHI 30 was associated with a significantly lower rate of daytime hypoglycemia (RR = 1.24; 95% CI, 1.08–1.43; P < 0.01; Table IV). Significantly fewer patients experienced a major hypoglycemic episode with BIAsp 30 compared with BHI 30 (P < 0.05; Figure 2), while rates of minor hypoglycemia were not significantly different between treatments (Table IV). Volume 31 Number 8
August 2009 Table II. Demographic information for the randomized, controlled clinical trials included in the meta-analysis. Trial ID 03819 (n = 187)
108812 (n = 93)
1234 (n = 180)
13539 (n = 428)
1394 (n = 292)
14668 (n = 160)
95 (50.8) 92 (49.2)
39 (41.9) 54 (58.1)
79 (43.9) 101 (56.1)
177 (41.4) 251 (58.6)
125 (42.8) 167 (57.2)
48 (30.0) 112 (70.0)
185 (98.9) 1 (0.5) 1 (0.5) – –
68 (73.1) 10 (10.8) 1 (1.1) 11 (11.8) 3 (3.2)
112 (62.2) – 64 (35.6) – 4 (2.2)
– – 428 (100) – –
290 (99.3) 1 (0.3) 1 (0.3) – –
154 (96.3) 5 (3.1) 1 (0.6) – –
Weight, mean (SD), kg 79.2 (13.0)
85.4 (13.3)
77.7 (15.7)
60.0 (9.8)
Age, mean (SD), y
56.9 (11.5) 60.3 (12.0)
Sex, no. (%) Female Male Race, no. (%) White Black Asian/PI Hispanic Other
63.2 (8.6)
83.6 (14.6) 86.2 (14.5)
60.9 (10.0) 65.8 (10.2)
62.9 (9.1)
1536 (n = 195) 98 (50.3) 97 (49.7) – – 195 (100) – –
3002 (n = 36) 16 (44.4) 20 (55.6) – – 36 (100) – –
3006 (n = 103)
46 (44.7) 713 (42.6) 57 (55.3) 961 (57.4) – – 103 (100) – –
66.6 (10.0) 63.7 (10.0) 63.5 (10.0) 57.8 (8.7)
Total* (N = 1674)
809 (48.3) 17 (1.0) 830 (49.6) 11 (0.7) 7 (0.4) –
53.5 (10.7) 53.8 (10.6) 61.0 (10.6)
PI = Pacific islander. *Patients with type 2 diabetes mellitus randomized and exposed to treatment.
J.A. Davidson et al.
1645
Clinical Therapeutics
Table III. Summary of hypoglycemic episodes in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30) (N = 1674). BIAsp 30 (n = 1150)* Episode Overall hypoglycemia† Major hypoglycemia Minor hypoglycemia Daytime hypoglycemia‡§ Nocturnal hypoglycemia§=
BHI 30 (n = 864)*
No. (%)
Episodes
No. (%)
Episodes
551 (47.9) 13 (1.1) 350 (30.4) 533 (46.3) 127 (11.0)
3205 18 2047 2782 369
382 (44.2) 20 (2.3) 203 (23.5) 343 (39.7) 147 (17.0)
2234 29 1234 1807 404
*Patients exposed to study drug (including 340 who received both in crossover studies). † Major, minor, and symptoms-only hypoglycemia. ‡ 6:01 AM to 11:59 PM. § Due to missing information, some hypoglycemic episodes (~2.5%) could not be assigned to either daytime or nocturnal. =12:00 AM to 6:00 AM.
Table IV. Meta-analyses of secondary end points in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30); rate ratios and treatment differences were calculated by the fixed-effects model. Variable
Rate Ratio* or Treatment Difference†
P
Heterogeneity (I2), %
Overall hypoglycemia‡ Minor hypoglycemia Daytime hypoglycemia§ HbA1c FPG PPG increment = Weight Insulin dose¶
1.08 (0.94 to 1.24)* 0.98 (0.81 to 1.20)* 1.24 (1.08 to 1.43)* 0.04 (–0.02 to 0.10)† 0.63 (0.31 to 0.95)† –0.31 (–0.49 to –0.07)† –0.07 (–0.24 to 0.09)† 0.01 (0.01 to 0.02)†
0.27 0.87 <0.01 0.25 <0.01 <0.01 0.40 <0.01
23 0 49 0 0 0 58 35
HbA1c = glycosylated hemoglobin; FPG = fasting plasma glucose; PPG = postprandial plasma glucose. * BIAsp 30/BHI 30. † BIAsp 30 – BHI 30. ‡ Major, minor, and symptoms-only hypoglycemia. § 6:01 AM to 11:59 PM. =Averaged over breakfast, lunch, and dinner. ¶ Only data for the first treatment period were included for the 2 crossover trials.
A sensitivity analysis, in which open-label studies were excluded from the meta-analysis of nocturnal hypoglycemia, yielded a similar result (RR = 0.41; 95% CI, 0.29–0.59; P < 0.01; I 2 = 34%) to that obtained when all trials were included (Figure 1). 1646
Glycemic Control All 9 trials provided HbA1c data for the metaanalysis. Three trials (1353,9 1466,8 and 1536) were excluded from the meta-analyses of PPG increment and FPG, as these data were not collected. Mean baseVolume 31 Number 8
J.A. Davidson et al.
Trial
Rate Ratio (95% CI)
P
03819
0.57 (0.20–1.58)
0.28
108812
0.89 (0.25–3.16)
0.86
1234
0.44 (0.22–0.89)
0.02
13539
1.03 (0.42–2.53)
0.95
1394
1.03 (0.38–2.76)
0.96
14668
0.33 (0.21–0.51)
0.01
1536
0.44 (0.11–1.47)
0.17
3002
1.05 (0.11–10.09)
0.97
3006
2.43 (0.31–18.90)
0.39
Overall*
0.50 (0.38–0.67)
<0.01
Test of Heterogeneity: I2 = 32% 0.1 0.2
1
Favors BIAsp 30
10
20
Favors BHI 30
Figure 1. Meta-analysis of nocturnal hypoglycemia in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30). *Data for all trials combined (fixed-effects estimate).
line and end-of-treatment values for HbA1c, FPG, and PPG increment are presented in Table V. BIAsp 30 was associated with reduced PPG increment (averaged over breakfast, lunch, and dinner) by a significantly greater extent than was BHI 30 (P < 0.01; Table IV), whereas BHI 30 treatment was associated with a significantly larger reduction in FPG (P < 0.01). However, no significant treatment difference was found for HbA1c.
Weight Eight clinical trials provided weight data for the meta-analysis. One trial (14668) was excluded because data were only collected at baseline. For trial 1234, change in weight from baseline was determined separately for each period. Both BIAsp 30 and BHI 30 were associated with an increase in weight from baseAugust 2009
line (mean [SD] change of 0.2 [2.3] and 0.7 [2.1] kg, respectively), but no significant treatment difference was found (Table IV).
Insulin Dose All 9 trials provided total daily insulin dose data for the meta-analysis. A higher increase from baseline in total daily dose was observed for BIAsp 30 relative to BHI 30 (P < 0.01; Table IV). For BIAsp 30, total daily dose increased from 0.54 (0.28) U/kg at baseline to 0.61 (0.31) U/kg; for BHI 30, total daily dose increased from 0.58 (0.30) U/kg at baseline to 0.64 (0.32) U/kg.
Heterogeneity In all cases where heterogeneity was found across trials for a particular end point (ie, I 2 >0%), similar 1647
Clinical Therapeutics
Trial
Odds Ratio (95% CI)
P
03819
0.50 (0.12–1.98)
0.32
1234
0.34 (0.01–8.28)
0.50
13539
0.57 (0.16–2.00)
0.38
1394
0.53 (0.03–8.63)
0.66
14668
0.31 (0.06–1.54)
0.15
3002
0.25 (0.01–6.62)
0.41
Overall*
0.45 (0.22–0.93)
<0.05
Test of Heterogeneity: I2 = 0% 0.01
0.1
1
Favors BIAsp 30
10
100
Favors BHI 30
Figure 2. Meta-analysis of major hypoglycemia in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30). *Data for all trials combined (fixed-effects estimate).
overall estimates and 95% CI were obtained when the meta-analysis was repeated using a random-effects model (data not shown).
DISCUSSION Intensive insulin treatment, designed to reduce the risk of long-term diabetic complications, has been associated with an increased risk of hypoglycemia in patients with T2DM, with clinical consequences ranging from mild discomfort to coma and even death.22 Although considered less common in patients with T2DM, rates of hypoglycemia may approach those observed in patients with type 1 diabetes with increasing age, duration of diabetes, and length of insulin treatment.23 Fear of hypoglycemia can result in a reluctance to initiate insulin treatment, even though insulin is recog1648
nized as the most effective diabetes medication to lower glycemia.1 Moreover, cases have been reported where patients intentionally keep blood glucose levels above recommended targets to avoid hypoglycemia.3 It is important, therefore, to evaluate insulin products and/or treatment regimens that carry a lower risk of hypoglycemia and do not jeopardize glycemic efficacy. In this study, we compared the rate of hypoglycemia associated with BIAsp 30 and BHI 30 using a meta-analysis of 9 published and unpublished randomized, controlled trials in adults with T2DM. Four of these trials also included patients with type 1 diabetes, but there was an insufficient number (151 patients) to conduct a meaningful meta-analysis on this patient population. Our results, based on individual data for 1674 patients with T2DM, suggested that, Volume 31 Number 8
J.A. Davidson et al.
Table V. Summary of glycemic end points in patients with type 2 diabetes mellitus treated with biphasic insulin aspart 30 (BIAsp 30) or biphasic human insulin 30 (BHI 30). BIAsp 30
HbA1c, % Baseline Weeks 12 to 16 Change from baseline to weeks 12 to 16 FPG, mmol/L Baseline Weeks 12 to 16 Change from baseline to weeks 12 to 16 PPG increment, mmol/L* Baseline Weeks 12 to 16 Change from baseline to weeks 12 to 16
BHI 30
No. of Patients
Mean (SD)
No. of Patients
Mean (SD)
965 929
8.1 (1.4) 7.9 (1.2)
690 647
8.3 (1.4) 7.9 (1.2)
929
–0.2 (1.0)
647
–0.3 (1.0)
466 458
9.4 (3.2) 9.4 (2.7)
400 396
9.6 (3.7) 8.9 (3.5)
447
0 (3.6)
389
–0.8 (4.2)
446 446
2.98 (2.22) 2.19 (2.26)
386 391
3.27 (2.47) 2.71 (2.24)
421
–0.81 (2.43)
372
–0.53 (2.58)
HbA1c = glycosylated hemoglobin; FPG = fasting plasma glucose; PPG = postprandial plasma glucose. *Averaged over breakfast, lunch, and dinner.
although no statistically significant difference in the overall rate of hypoglycemia was found between treatments, BIAsp 30 was associated with a significantly lower rate of nocturnal hypoglycemia compared with BHI 30, but a significantly higher rate of daytime hypoglycemia. Hypoglycemia is unwelcome at any time of the day but may be a particular problem at night, because a sleeping patient may be unaware of symptoms that would have otherwise prompted corrective action. Whereas no significant difference was found in the frequency of minor hypoglycemia associated with BIAsp 30 or BHI 30, significantly fewer patients on BIAsp 30 experienced a major hypoglycemic event. Because major hypoglycemia can be potentially life threatening (and incur high health care costs), reducing the risk for such events is essential for both the patient and health care industry.24 In terms of glycemic control, the meta-analysis found that BIAsp 30 was associated with reduced PPG increment averaged over the 3 main meals to a signifiAugust 2009
cantly greater extent than was BHI 30, whereas a significantly larger reduction in FPG was found for BHI 30. However, overall metabolic control (as measured by HbA1c) did not differ significantly between treatment groups. Therefore, the reduced risk for nocturnal and major hypoglycemia associated with BIAsp 30 was not achieved at the expense of overall glycemic control. It is also noteworthy that the comparable overall glycemic control provided by BIAsp 30 and BHI 30 was obtained with similar total daily doses of insulin, and that both treatments resulted in similar weight gain. For both treatment groups, HbA1c following 12 to 16 weeks of treatment was ~7.9%. This is above ADA- and EASD-recommended HbA1c targets of <7%25 and may reflect the fact that none of the trials in the meta-analysis were of a treat-to-target design. Due to the relatively short treatment duration of the majority of clinical trials in the meta-analysis (12 to 16 weeks) and the low number of major hypoglycemic episodes reported in these studies, our findings for HbA1c and major hypoglycemia need to be 1649
Clinical Therapeutics confirmed by a larger, properly powered, randomized, controlled trial of sufficient length. The meta-analysis confirmed previous findings that BIAsp 30 provides significantly better postprandial glycemic control than does BHI 30.9–11,17 The improved mealtime coverage has been attributed to the faster onset and higher plasma insulin levels attained by the soluble insulin aspart component of BIAsp 30 compared with the regular human insulin constituent of BHI 30.4 Shortacting human insulin has a longer pharmacodynamic action profile than does insulin aspart,26 with this characteristic likely to account, at least in part, for the significantly higher rate of nocturnal hypoglycemia with BHI 30 in this study. The observed peaking action of the intermediate-acting NPH insulin component of BHI 30 may also contribute to the higher rate of nocturnal hypoglycemia and lower FPG levels observed for this premixed insulin.27 As higher prandial plasma insulin levels are attained with BIAsp 30 than BHI 30,4 it could be speculated that the higher rate of daytime hypoglycemia associated with this insulin premix may be partially attributed to occasions in which there is inadequate carbohydrate intake with the mealtime dose. Low to moderate heterogeneity was found across trials for some of the end points analyzed (weight, insulin dose, overall hypoglycemia, daytime hypoglycemia, and nocturnal hypoglycemia). However, in all cases, the same conclusions were reached for the end point regardless of whether a fixed- or random-effects model was used for the meta-analysis, thereby supporting the robustness of the results. To test for any potential bias in the reporting of hypoglycemic events for open-label trials, a sensitivity analysis was performed for the primary end point whereby only trials that were blinded to treatment allocation were included in the meta-analysis of nocturnal hypoglycemia. Near-identical results were found to those observed for the full meta-analysis dataset, indicating that reporting bias was unlikely.
CONCLUSION This meta-analysis found BIAsp 30 to be associated with a significantly lower rate of nocturnal and major hypoglycemia, but a significantly increased risk for daytime hypoglycemia, compared with BHI 30 at a similar level of HbA1c in patients with T2DM.
ACKNOWLEDGMENTS The authors wish to thank the study coordinators, nurses, physicians, and patients who participated in 1650
the clinical trials included in this meta-analysis. We also thank Paul G. Drake, PhD, Novo Nordisk A/S, who assisted in the preparation of this manuscript. Dr. Davidson has been a consultant (speaker, advisory board member, clinical trials) for Novo Nordisk. Dr. Liebl has received honoraria from Novo Nordisk for scientific presentations and participating at advisory board meetings, and has received financial help with clinical studies from Novo Nordisk. Dr. Christiansen has received research grants and speakers’ fees from Novo Nordisk. Dr. Fulcher has received honoraria from Novo Nordisk for speaking at educational activities and participating at advisory board meetings. Dr. Ligthelm has been a member of Novo Nordisk advisory boards and has been a speaker at Novo Nordisk meetings. Dr. Gylvin owns shares in Novo Nordisk. Dr. Kawamori has received honoraria for lectures cosponsored by Novo Nordisk.
REFERENCES 1. Nathan DM, Buse JB, Davidson MB, et al. Management of hyperglycemia in type 2 diabetes: A consensus algorithm for the initiation and adjustment of therapy: A consensus statement from the American Diabetes Association and the European Association for the Study of Diabetes [published correction appears in Diabetes Care. 2006;49:2816–2818]. Diabetes Care. 2006;29:1963–1972. 2. Jellinger PS, Davidson JA, Blonde L, et al, for the ACE/ AACE Diabetes Road Map Task Force. Road maps to achieve glycemic control in type 2 diabetes mellitus: ACE/ AACE Diabetes Road Map Task Force. Endocr Pract. 2007;13:260–268. 3. Wild D, von Maltzahn R, Brohan E, et al. A critical review of the literature on fear of hypoglycemia in diabetes: Implications for diabetes management and patient education. Patient Educ Couns. 2007;68:10–15. 4. Garber AJ, Ligthelm R, Christiansen JS, Liebl A. Premixed insulin treatment for type 2 diabetes: Analogue or human? Diabetes Obes Metab. 2007;9:630–639. 5. Halimi S, Raskin P, Liebl A, et al. Efficacy of biphasic insulin aspart in patients with type 2 diabetes. Clin Ther. 2005;27(Suppl B):S57–S74. 6. Gough SC, Tibaldi J. Biphasic insulin aspart in type 2 diabetes mellitus: An evidence-based medicine review. Clin Drug Investig. 2007;27:299–324. 7. Boehm BO, Vaz JA, Brøndsted L, Home PD. Long-term efficacy and safety of biphasic insulin aspart in patients with type 2 diabetes. Eur J Intern Med. 2004;15:496–502. 8. McNally PG, Dean JD, Morris AD, et al. Using continuous glucose monitoring to measure the frequency of low glucose values when using biphasic insulin aspart 30 com-
Volume 31 Number 8
J.A. Davidson et al.
9.
10.
11.
12.
13.
14.
15.
16.
pared with biphasic human insulin 30: A double-blind crossover study in individuals with type 2 diabetes. Diabetes Care. 2007;30:1044–1048. Iwamoto Y. A randomised, multicentre trial of biphasic insulin aspart versus biphasic human insulin in type 2 diabetes. Diabetologia. 2003; 46:A270. McSorley PT, Bell PM, Jacobsen LV, et al. Twice-daily biphasic insulin aspart 30 versus biphasic human insulin 30: A double-blind crossover study in adults with type 2 diabetes mellitus. Clin Ther. 2002;24:530–539. Hermansen K, Colombo M, Storgaard H, et al. Improved postprandial glycemic control with biphasic insulin aspart relative to biphasic insulin lispro and biphasic human insulin in patients with type 2 diabetes. Diabetes Care. 2002;25:883– 888. Kilo C, Mezitis N, Jain R, et al. Starting patients with type 2 diabetes on insulin therapy using oncedaily injections of biphasic insulin aspart 70/30, biphasic human insulin 70/30, or NPH insulin in combination with metformin. J Diabetes Complications. 2003;17:307–313. Wright AD, Cull CA, Macleod KM, et al. Hypoglycemia in type 2 diabetic patients randomized to and maintained on monotherapy with diet, sulfonylurea, metformin, or insulin for 6 years from diagnosis: UKPDS73. J Diabetes Complications. 2006;20:395–401. Whitehead A, ed. Meta-Analysis of Controlled Clinical Trials. Chichester, UK: John Wiley & Sons; 2002. The Cochrane Collaboration. The Cochrane Policy Manual. Issue 3. http://www.cochrane.org/admin/ manual.htm. Accessed July 16, 2009. ICH Harmonised Tripartite Guideline: Guideline for Good Clinical Practice E6(R1). June 1996. http:// www.emea.europa.eu/pdf s/ human/ich/013595en.pdf. Accessed January 16, 2009.
August 2009
17. Kapitza C, Rave K, Ostrowski K, et al. Reduced postprandial glycaemic excursion with biphasic insulin Aspart 30 injected immediately before a meal. Diabet Med. 2004;21:500– 501. 18. Abrahamian H, Ludvik B, Schernthaner G, et al. Improvement of glucose tolerance in type 2 diabetic patients: Traditional vs modern insulin regimens (results from the Austrian Biaspart Study). Horm Metab Res. 2005;37:684–689. 19. Boehm BO, Home PD, Behrend C, et al. Premixed insulin aspart 30 vs premixed human insulin 30/70 twice daily: A randomized trial in type 1 and type 2 diabetic patients [published correction appears in Diabet Med. 2002;19:797]. Diabet Med. 2002; 19:393–399. 20. International Conference on Harmonisation (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Harmonised Tripartite Guideline. Statistical principles for clinical trials: E9. http://www.ich.org/LOB/ media/MEDIA485.pdf. Accessed July 9, 2009. 21. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327: 557–560. 22. Cryer PE, Davis SN, Shamoon H. Hypoglycemia in diabetes. Diabetes Care. 2003;26:1902–1912.
23. UK Hypoglycaemia Study Group, ROYAUME-UNI. Risk of hypoglycaemia in types 1 and 2 diabetes: Effects of treatment modalities and their duration. Diabetologia. 2007;50: 1140–1147. 24. Amiel SA, Dixon T, Mann R, Jameson K. Hypoglycaemia in type 2 diabetes. Diabet Med. 2008;25:245–254. 25. Nathan DM, Buse JB, Davidson MB, et al, for the European Association for the Study of Diabetes. 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. 2009;32:193–203. 26. Mudaliar SR, Lindberg FA, Joyce M, et al. Insulin aspart (B28 asp-insulin): A fast-acting analog of human insulin: Absorption kinetics and action profile compared with regular human insulin in healthy nondiabetic subjects. Diabetes Care. 1999;22: 1501–1506. 27. Heise T, Heinemann L, Hövelmann U, et al. Biphasic insulin aspart 30/70 (BIAsp 30): Pharmacokinetics (PK) and pharmacodynamics (PD) in comparison with once-daily biphasic human insulin and basalbolus therapy [Published online ahead of print June 1, 2009]. Diabetes Care.
Address correspondence to: Paul Brown, CStat, Department of Biostatistics, Novo Nordisk A/S, Vantaarnsvej 114, 2860 Soeborg, Denmark. E-mail: [email protected] 1651