A Pooled Analysis of Treat-To-Target Trials

Current Medical Research & Opinion Vol. 26, No. 7, 2010, 1621–1628

Original article Starting insulin therapy with basal insulin analog or premix insulin analog in T2DM: a pooled analysis of treat-to-target trials

Vivian Fonseca Abstract Tulane University Health Sciences Center, New Orleans, LA, USA Objective: Although the choice of starting insulin for people with type 2 diabetes mellitus (T2DM) is often a basal or Jaime Davidson premix insulin analog, there is little evidence to base this decision on. This analysis aimed to determine if UT Southwestern Medical School & Medical Center, measurable clinical characteristics prior to starting insulin could predict differences in improved glycemic Dallas, TX, USA control between these options. Philip Home Newcastle University, Newcastle upon Tyne, UK Research design and methods: James Snyder A thorough literature search was performed for treat-to-target studies in insulin-naı¨ve individuals with T2DM Las Vegas, NV, USA treated with biphasic insulin aspart (BIAsp 30), a basal insulin analog (insulin glargine or insulin detemir) or NPH insulin regimens once or twice daily plus oral glucose-lowering drugs (OGLDs). Patient data were Paul Jellinger thorised users can download, pooled and mean baseline age, diabetes duration,le copy gender, for body personal mass index (BMI), use HbA 1c, fasting plasma Center For Diabetes & Endocrine Care, Hollywood, glucose (FPG), average postprandial plasma glucose over three meals (PPG) and bedtime PG were For personal use only. CA, USA investigated for prediction of improved HbA 1c, FPG or PPG. Statistical analyses employed a linear mixed Anders Dyhr Toft model with insulin type, OGLD, time and time-squared as fixed effects and patient and trial as random Novo Nordisk A/S, Virum, Denmark effects. Anthony Barnett Results: University of Birmingham and Heart of England NHS Foundation Trust, Birmingham, UK Baseline age ( p ¼ 0.022) and bedtime PG ( p ¼ 0.036) were inter-related predictors of HbA 1c. In older individuals with higher bedtime PG, BIAsp 30 appeared to be more beneficial. In contrast, basal insulin Copyright ©appeared 2010 to be a better Informa choice in younger individualsUK Limited with lower bedtime PG. For FPG, BMI ( p ¼ 0.011) and Address for correspondence: post-breakfast PG ( p ¼ 0.042) were identified as predictors. Basal insulin appeared to achieve better FPG in Prof. Vivian Fonseca, Tulane University Health patients with lower BMI and higher post-breakfast PG, while BIAsp 30 appeared to be better in patients with Sciences Center, 1430 Tulane Avenue, New Orleans, Not for Salehigher BMIor and Commercial lower post-breakfast PG. Age ( Distributionp ¼ 0.0347) was the only baseline characteristic associated LA 70112, USA. display, view and print a sing Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 Tel.: þ 1 504 988 4023; Fax: þ1 504 988Unauthorized 6271; with use differences prohibited. in average PPG: Au mean PPG was similar between regimens in younger people, but BIAsp 30 [email protected] achieved better PPG results in older persons. Minor hypoglycemia was reported by 56% of BIAsp 30- and 45% of basal-treated individuals. The major limitation of the study was that only Novo Nordisk trials were included in the analysis as access to individual patient data was required. As the trials were fairly Key words: heterogeneous a strict methodology was used to minimize potential confounding factors. Blood glucose – Insulin aspart – Insulin detemir – Insulin glargine – Pooled analysis – Type 2 diabetes Conclusions: Accepted: 9 April 2010; published online: 29 April 2010 Premix analog rather than basal insulin may be an appropriate choice to target HbA 1c values in older Citation: Curr Med Res Opin 2010; 26:1621–28 individuals and those with higher bedtime PG, while basal insulin may be more appropriate to target FPG in patients with lower BMI and higher post-breakfast PG.

Introduction For individuals with T2DM who are unable to control their blood glucose with maximal dosing of oral glucose-lowering drugs (OGLDs), the choice for starting

! 2010 Informa UK Ltd www.cmrojournal.com Pooled analysis of basal versus premix insulin analog Fonseca et al. 1621 Current Medical Research & Opinion Volume 26, Number 7 July 2010

insulin is commonly between a basal or premix insulin most benefit from one or the other of the insulin types regimen1,2, although a mealtime þ basal approach can using characteristics derived from an evidence-based anal- also be used3. Given that this is an important choice for ysis of the published studies. Therefore, the current pooled patients and their physicians to make, there is a relative analysis was performed to identify whether differences in lack of evidence-based information available about which patients’ baseline criteria, including HbA1c, fasting plasma regimen is more effective when starting insulin in terms of glucose (FPG), post-prandial glucose (PPG), age, gender baseline clinical characteristics. Accordingly, lifestyle fac- or duration of diabetes, could be used to predict differences tors, convenience and physician preferences often play a in the clinical efficacy of the basal or premix insulin analog large part in determining the chosen regimen. For exam- regimens. ple, physicians may prefer premix insulin for elderly patients, those with vision problems or those with arthritis because they offer the ability to address both basal and prandial glucose control with fewer daily injections Patients and methods versus a multiple injection approach. A thorough literature search of PubMed was carried out A number of studies have compared basal and premix using the search terms ‘insulin apart’, ‘insulin lispro’, ‘insu- insulin analog regimens in combination with one or more lin detemir’, ‘insulin glargine’ and ‘insulin naı¨ve’. OGLDs using a treat-to-target approach3–11. Overall, Abstracts were reviewed to identify treat-to-target studies these studies demonstrated that while both basal-only comparing insulin initiation with premix or basal insulin. and premix insulin regimens significantly improve The study then included all Novo Nordisk clinical trials in HbA1c levels, premix insulin may offer benefits in terms insulin-naı¨ve patients with T2DM who started treatment of the magnitude of HbA1c reduction. However, the more with once- or twice-daily BIAsp 30 according to a treat- intensive premix insulin regimens also tended to be asso- to-target protocol (biphasic insulin aspart, NovoMix 30, ciated with increased rates of minor hypoglycemic episodes Novo Nordisk A/S, Bagsværd, Denmark or NovoLog Mix and greater weight gain compared with basal regimens. 70/30 in the US, Novo Nordisk Inc., Princeton, New Furthermore, the OGLDs used together with the insulin Jersey, USA). In addition, trials of the basal insulin analogs may also have affected the rates of hypoglycemia and (insulin glargine [Lantus, Sanofi Aventis, Bridgewater, weight effects observed in the studies, particularly in New Jersey, USA] or insulin detemir [Levemir, Novo those studies in which insulins were combined with Nordisk A/S, Bagsværd, Denmark]) or NPH insulin once

For personal use only. sulfonylureas. or twice daily using a treat-to-target protocol were Given the differences between these two treatment included (Table 1)4–9. In all trials, insulin treatment was options in terms of the balance between clinical efficacy combined with one or more OGLDs. The literature search and the potential for unwelcome side effects, it would be also identified a number of non-Novo Nordisk studies in helpful to be able to identify those patients who would insulin-naı¨ve patients that examined the initiation of basal

Table 1. Key characteristics of the included trials.

Study Insulin treatment OGLD treatment Randomized Trial duration patients (n) (weeks)

Rosenstock et al., 2008 Detemir once or twice daily vs. Metformin, insulin secretagogue and 582 52 5 Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 80 sites (Europe and the USA) once-daily glargine (partici- alpha-glucosidase inhibitors in pants on once-daily detemir mono- or combination therapy may switch to twice daily) Hermansen et al., 2006 Twice-daily detemir vs. twice- Metformin, insulin secretagogue or 476 24 58 sites in 10 European countries6 daily NPH alpha-glucosidase inhibitors in mono- or combination therapy Philis-Tsimikas et al., 2006 Two arms once-daily detemir Continue current OGLD therapy. 504 20 91 sites (Europe and the USA)8 (morning and evening) vs. Alpha-glucosidase inhibitor mono- once-daily NPH therapy not allowed. TZDs may be taken if approved Raskin et al., 2005 Twice-daily BIAsp 30 vs. once- Metformin with and without TZDs 233 30 25 sites in the USA9 daily glargine Garber et al., 2006 Intensifying treatment starting on Continue prior OGLD regimen 100 16 (only first phase 7 13 sites in the USA once-daily BIAsp 30, if HbA1c is (enrolled) included) above 6.5 a second injection was added Raskin et al., 2009 Twice-daily BIAsp 30 and OGLDs Metformin and pioglitazone 200 24 73 centers in the USA4 vs. OGLDs only

NPH ¼ neutral protamine Hagedorn; OGLD ¼ oral glucose-lowering drug; TZD ¼ thiazolidinedione.

1622 Pooled analysis of basal versus premix insulin analog Fonseca et al. www.cmrojournal.com ! 2010 Informa UK Ltd Current Medical Research & Opinion Volume 26, Number 7 July 2010

insulin analogs or premix insulin using a treat-to-target Statistical modeling approach10–16. Due to the need for patient-level data the published reports did not include enough detail to be A two-round screening procedure was applied to assess included in the analysis. Requests were made for patient- which baseline characteristics had a significant effect on level data from the manufacturers of insulin lispro and glycemic endpoints and to compare any variation in effec- insulin glargine, but these data were not available. tiveness between the basal and premix insulin analog regimens. The parameters included in this analysis are listed Therefore, only Novo Nordisk trials were included in the in Table 3. In the first round, each baseline characteristic analysis. was studied individually to identify those in which the two Demographic characteristics of the participants in the insulin regimens had significantly different effects on a included studies are shown in Table 2. All randomized particular endpoint. All of those baseline characteristics insulin-naı¨ve participants who were exposed to at least so identified were analyzed simultaneously in the second one dose of trial medication were included in the pooled round using a backwards selection strategy. Baseline char- analysis. Due to differences in treatment regimens between acteristics that were found to have a significant impact on the studies, participants were separated into three groups the endpoints in the second screening round were included defined according to their insulin treatment: (1) Basal: in the final analysis. participants who were treated with insulin detemir During the screening procedure and the final statistical (detemir, once or twice daily as the only insulin) or insulin analysis, continuous endpoints were analyzed using analy- glargine (glargine, once daily) plus OGLDs, (2) Premix: sis of variance (ANOVA). All of the models included the participants who were treated with BIAsp 30 once or following variables: twice daily plus OGLDs, (3) NPH: participants who (1) Insulin type as fixed effect were treated with NPH insulin only once or twice daily (2) OGLDs as fixed effect plus OGLDs. The grouping of the insulin treatments was (3) Time (as days from randomization) and time-squared necessary to analyze overall treatment effects rather than (to allow for non-linear relationships) as covariates individual trial effects. In order to increase the robustness (4) Trial as random effect of the statistical model, participants treated with NPH (5) Subject as random effect were included as a supportive arm to provide information (6) Baseline value of endpoint as covariate (i.e., if HbA1c on non-treatment-specific relationships between end- is analyzed, the model will contain baseline HbA1c) points and baseline characteristics. The target endpoints were screened and analyzed in a sim- For personal use only. No comparison between NPH and the other two arms ilar manner as the continuous endpoints, but with a mar- was performed due to differences in pharmacokinetic prop- ginal logistic regression model instead of an ANOVA erties and within-person variability in pharmacodynamic model with random subject and trial effects. The model endpoints between NPH and the insulin analog arms. elements were similar to the continuous counterpart but with trial included as fixed effect. The parameters in the model were estimated using generalized estimating equa- Influencing factors tions (GEE) methodology, with a compound symmetry structure on the correlation matrix describing the correla- Different OGLD mono- and combination therapies were tion between the multiple endpoint values from the same included in the analyzed trials, which could have affected patient. Baseline characteristics were included in the both glycemic outcomes, thus confounding the analysis. model in the different stages of analysis in the same Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 Duration of trial participation was also considered as a manner as for the continuous case. The logistic regression confounding factor, because the longer an individual par- model was fitted using a logit link function. ticipates in a treat-to-target trial, the more likely it is that therapy will be optimized. Therefore, OGLD therapy and the duration of trial exposure were considered to be possi- ble confounding factors and were controlled for in the data Results analysis. Due to differences in concomitant OGLD treatment regimens between the studies, participants were Absolute HbA1c separated into four groups defined according to their The first round of the screening process identified four insulin treatment: (1) metformin monotherapy, (2) thiazo- baseline characteristics that were potentially or actually lidinedione monotherapy, (3) secretagogue monotherapy, associated with treatment differences between the basal (4) combination therapy of two or more OGLDs. The and biphasic insulin regimens (Table 4a). Analysis of grouping of concomitant OGLD treatments was these four baseline characteristics in the second-round necessary to analyze overall treatment effects rather than screen identified age (p ¼ 0.022) and bedtime PG individual trial effects. (p ¼ 0.036) as having a significant impact on the HbA1c

! 2010 Informa UK Ltd www.cmrojournal.com Pooled analysis of basal versus premix insulin analog Fonseca et al. 1623 Current Medical Research & Opinion Volume 26, Number 7 July 2010

Table 3. Parameters included in the analysis. 1)

4 Baseline characteristics Endpoints Met/pio

HbA1c (%) HbA1c (%) FPG (mmol/L) FPG (mmol/L) Average PPG (mmol/L) over Average PPG (mmol/L) over Raskin, 2009

30 three meals three meals BIAsp Post-breakfast PG (mmol/L) Proportion of participants Post-dinner PG (mmol/L) reaching HbA1c 57 (%) Bedtime PG (mmol/L) Proportion of participants 7 Pre-dinner PG (mmol/L) reaching HbA1c 6.5 (%) Diabetes duration (years) Proportion of participants

30 Gender (male/female) reaching FPG56.1 (mmol/L) Ethnicity (Caucasian/not Proportion of participants Garber, 2006 Caucasian) reaching average PPG510 ALT (U/L) (mmol/L) Age (years) Proportion of participants BMI (kg/m2) reaching average PPG57.8

5.3 34.2 (6.7) 32.4 (5.2) 33.4 (5.7) (mmol/L) 9 Glargine BIAsp ALT ¼ alanine aminotransferase; BMI ¼ body mass index; FPG ¼ fasting plasma glucose; HbA1c ¼ glycated hemoglobin; PG ¼ plasma glucose; PPG ¼ postprandial plasma glucose. Raskin, 2005 30

BIAsp endpoint; accordingly, these were included in the final model (Table 4b). Therefore, baseline HbA1c, FPG, average PPG (mmol/L), post-breakfast PG, post-dinner PG, pre-dinner PG, diabetes duration, gender, ethnicity, ala- NPH

insulin nine aminotransferase (ALT) or BMI were not identified Evening 8 by the screening process as being associated with treatment

pioglitazone. differences between basal and biphasic insulin regimens. ¼ The model-estimated mean HbA1c for participants insulin detemir Evening treated with basal insulin and BIAsp 30 for different

For personal use only. values of age and bedtime PG can be seen in Figure 1.

Philis-Tsimikas, 2006 Statistical testing was not performed on the different data points presented in Figure 1; however, the data sug- insulin detemir

Morning gest that basal insulin may be a better initiation option in younger patients with lower pre-insulin bedtime PG, while BIAsp 30 may be a better option in older patients with neutral protamine Hagedorn, pio 6

¼ higher pre-insulin bedtime PG. Thus, at bedtime PG of NPH insulin 178mg/dL (9.9mmol/L), BIAsp 30 was better for those over 35 years old, whereas at age 45 years, basal insulin was better if pre-insulin bedtime PG was below 130mg/dL

metformin, NPH (7.2mmol/L). Figure 1 also suggests that BIAsp 30 may be a Hermansen, 2006 ¼ Insulin detemir Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 better option in patients aged over 55 years, whatever their bedtime PG. 5 Insulin

glargine Fasting plasma glucose male/female, met ¼ The screening process identified two baseline characteristics: BMI (basal insulin 0.054, BIAsp 30 0.010, difference 2 Rosenstock, 2008 291 291 237 338 1650.043kg/m 169; p ¼ 1640.011) and 117 post-breakfast 116 PG 100 (basal 102 98 Insulin detemir 9.1 (6.1) 9.1 (6.4) 9.6 (6.6) 9.8 (6.2) Not reported 9.5 (5.9) 8.9 (4.8) 11.1 (7.1) 9.2 (6.2) 8.3 (5.6)

30.6 (4.8) 30.5 (4.6) 28.9 (3.6) 29.0 (3.6) 29.8 (5.0) 29.7 (5.1)0.009, 30.4 (4.8) BIAsp 31.5 (5.5) 31.4 30 0.032, difference 0.756mg/dL (0.042mmol/L); p ¼ 0.042), which predicted FPG. Therefore, HbA1c, FPG, average PPG over three meals,

2 post-dinner PG (mmol/L), bedtime PG (mmol/L), pre- biphasic insulin aspart, M/F Demographic characteristics of the participants in the included studies. ¼ , % 8.64 (0.78) 8.62 (0.77) 8.61 (0.78) 8.51 (0.76) 9.08 (0.97) 8.88dinner (0.95) 9.15 (1.0) PG 9.7 (1.5) (mmol/L), 9.8 (1.4) diabetes 8.6 (0.8) duration, 8.1 (1.0) 8.1 gender, (1.0) ethnicity, 1c duration, years ALT and age were not associated with differences in FPG Mean age, years 58.4 (10.2) 59.4 (9.6) 61.3 (9.1) 60.4 (9.3) 58.3 (10.4) 58.7 (10.2) 58.4 (11.0) 52.6 (10.6) 52.3 (9.8) 56.7 (11.5) 53.4 (10.1) 54.2 (10. N HbA BIAsp Sex, % M/FWeight, kgBMI, kg/m 57/43 87.4 (16.6) 87.4 59/41 (17.4) 82.7 (13.3) 49/51 82.5 (14.2) 57/43 59/41 Not reported 54/46 57/43 90.6 (18.8) 53/47 89.9 (19.0) Not reported 56/44 Not reported 50/50 46/54 38/62 Data are shown as value (standard deviation). Diabetes

Table 2. outcomes with the basal and biphasic insulin treatments.

1624 Pooled analysis of basal versus premix insulin analog Fonseca et al. www.cmrojournal.com ! 2010 Informa UK Ltd Current Medical Research & Opinion Volume 26, Number 7 July 2010

Table 4. Baseline characteristics associated with differences in absolute HbA1c following basal or premix therapy: a) first-round screen, b) second-round screen.

Variable Basal estimate BIAsp 30 estimate Basal – BIAsp 30 estimate Value SE Value SE Value 95% CI p-value

a) First-round screen Age (years) 0.005 0.002 0.014 0.004 0.009 0.001; 0.018 0.0650 Average PPG (mmol/L) 0.008 0.008 0.025 0.012 0.033 0.006; 0.059 0.0165 Bedtime PG (mmol/L) 0.011 0.007 0.031 0.010 0.041 0.018; 0.064 0.0005 Post-dinner PG (mmol/L) 0.004 0.007 0.018 0.010 0.022 0.001; 0.045 0.0588 Basal estimate BIAsp 30 estimate Basal – BIAsp 30 estimate Value SE Value SE Value SE p-value b) Second-round screen Age (years) 0.004 0.002 0.015 0.004 0.011 0.005 0.0224 Average PPG (mmol/L) 0.014 0.014 0.008 0.027 0.007 0.029 0.8252 Bedtime PG (mmol/L) 0.010 0.009 0.045 0.016 0.054 0.019 0.0036 Post-dinner PG (mmol/L) 0.012 0.012 0.009 0.022 0.021 0.025 0.4066

CI ¼ confidence interval; BIAsp ¼ biphasic insulin aspart; PG ¼ plasma glucose; PPG ¼ postprandial plasma glucose; SE ¼ standard error.

15.1 14.10

–0.50 Favors basal 12.6 11.85 insulin Favors BIAsp 30

–0.25 0.5 9.60 0.4 10.1 0.00 0.3 0.2 0.1

For personal use only. 0.0

–0.1

Bedtime PG (mmol/L) 7.6 7.35 –0.2

Post-breakfast PG (mmol/L) Post-breakfast Favors 0.25 BIAsp 30 Favorsinsulin basal

5.1 5.10 15.1 21.1 27.1 33.1 39.1 20.1 35.1 50.1 65.1 80.1 2 Age (years) BMI (kg/m )

Figure 1. Model-determined treatment difference in predicted HbA1c (%) Figure 2. Treatment difference in fasting plasma glucose (mmol/L) (BIAsp (BIAsp 30 – basal insulin) according to age and bedtime PG. 30 – basal insulin) according to BMI and post-breakfast PG. Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 Figure 2 shows the model-based mean FPG for BIAsp 30 Post-prandial plasma glucose and basal insulin for different values of BMI and post- breakfast PG. Statistical testing was not performed on Age (basal 0.021, BIAsp 30 0.005, difference 0.026 years, the different data points presented in Figure 2. However, p ¼ 0.0347) was the only baseline characteristic associated these data suggest that basal insulin is likely to achieve with differences in average PPG with BIAsp 30 and basal better FPG in patients with lower BMI and higher post- insulin during the screening process. Therefore, HbA1c, breakfast PG; for example, a patient with a BMI of FPG, average PPG over three meals, post-breakfast PG, 25.0kg/m2 and a post-breakfast PG of 173mg/dL post-dinner PG, bedtime PG, pre-dinner PG, diabetes (9.6mmol/L) would have a 5.4mg/dL (0.3mmol/L) FPG duration, gender, ethnicity, ALT and BMI were not asso- advantage if started on a basal insulin rather than BIAsp ciated with treatment differences in PPG between basal 30. In contrast, a patient with a BMI of 36.0kg/m2 and a insulin and biphasic insulin. Model-estimated mean aver- post-breakfast PG of 132mg/dL (7.4mmol/L) would have a age PPG (over three meals) for different values of age is 3.6mg/dL (0.2mmol/L) FPG advantage if started on BIAsp presented in Figure 3. The model demonstrated that there 30 rather than basal insulin. were fewer differences in mean average PPG between

! 2010 Informa UK Ltd www.cmrojournal.com Pooled analysis of basal versus premix insulin analog Fonseca et al. 1625 Current Medical Research & Opinion Volume 26, Number 7 July 2010

treatments in younger patients; however, in older patients, risk of minor hypoglycemia. Accordingly, the model sug- average PPG values were lower with BIAsp 30 than with gests that younger people with lower bedtime PG starting basal insulin. For example, Figure 3 suggests that at 40 insulin therapy should use a basal analog. However, some years of age, initiating insulin with BIAsp 30 would lead caution is required with the latter conclusion, given that to an approximately 10.8mg/dL (0.6mmol/L) PPG advan- the clinical trials on which the model was based included tage compared with basal insulin, while at age 70, data from predominantly older individuals. These results this advantage would be approximately 21.6mg/dL are in agreement with the major studies on which the (1.2mmol/L) favoring BIAsp 30. model is based. Thus, the Initiate Insulin by Aggressive Titration and Education (INITIATE) study showed that beginning insu- Hypoglycemia lin with twice-daily BIAsp 30 was associated with improved HbA reduction compared with once-daily Major hypoglycemic episodes were uncommon with either 1c insulin glargine (2.8% and 2.4%, respectively; insulin, with a rate of approximately six episodes per 100 p50.05)9. Other trials, such as the Treating to Target in exposure-years (Table 5). The rate of minor hypoglycemic Type Two diabetes (4-T) study, reported similar findings episodes was approximately 40% lower with basal insulin in terms of HbA reductions when comparing twice-daily than BIAsp 30 (Table 5). However, rates of hypoglycemic 1c BIAsp 30 and once-daily basal insulin3. Similarly, studies episodes may be affected by the OGLD treatment that the not included in this model using twice-daily biphasic insu- insulins were combined with. lin lispro plus metformin have also demonstrated improved HbA1c reductions compared with once-daily insulin glargine plus metformin13,14. One finding of the INITIATE Discussion study was that participants with a higher baseline HbA1c (48.5%) did particularly well on BIAsp 30 compared with The results from this study suggest that starting BIAsp 30 9 insulin glargine ; however, our data can neither corrobo- rather than basal insulin would be a more appropriate rate nor contradict this finding. choice to lower HbA1c values in older individuals and This study also suggested that basal insulin might be a those with higher bedtime PG, although with a higher better option to target FPG in those patients with lower BMI and higher post-breakfast PG. It should be noted, 20 19 however, that the study participants who received basal For personal use only. 18 insulin were usually injected at bedtime, while those trea- 17 16 ted with BIAsp 30 generally received their last injection 15 with their evening meal. Given that studies of NPH versus 14 13 long-acting insulin analogs injected in the evening show 12 little difference in FPG6,11, the finding that basal insulin 11 10 provides better FPG control than BIAsp 30 is perhaps not 9 surprising. While the model suggests that BIAsp 30 would 8 be appropriate for patients with higher BMI and lower

Average PPG (mmol/L) Average 7 6 post-breakfast PG, the majority of data included in the 5 4 model did not come from such people, so some caution 3 should be exercised when accepting this conclusion. The Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 20 30 4050 60 70 80 90 finding that basal insulin is more able to reduce FPG values Age (years) than biphasic insulin is consistent with some randomized, Basal fitted BIAsp 30 fitted 3,13 Basal obs BIAsp 30 obs prospective trials , but was not found in the INITIATE study9. However, target FPG values of 80–100mg/dL Figure 3. Predicted average postprandial plasma glucose (mmol/L) by age. (4.4–5.6mmol/L) were found in a greater proportion of

Table 5. Hypoglycemic episodes.

Basal insulin BIAsp 30 Classification N (%) ETRate N (%) ETRate

Major 23 2.1 47 752 0.06 7 2.4 8 138 0.06 Minor 509 45.2 3613 752 4.80 161 55.9 1151 138 8.32

BIAsp ¼ biphasic insulin aspart; N ¼ number of patients experiencing at least one hypoglycemic episode; % ¼ percentage of patients experiencing at least one hypoglycemic episode; E ¼ number of episodes; T ¼ exposure years; Rate ¼ rate of hypoglycemic episodes per exposure year.

1626 Pooled analysis of basal versus premix insulin analog Fonseca et al. www.cmrojournal.com ! 2010 Informa UK Ltd Current Medical Research & Opinion Volume 26, Number 7 July 2010

individuals treated with basal (57%) compared with bipha- basal insulin in terms of prandial control. In contrast, basal sic insulin (36%)9. insulin is usually is administered at bedtime rather than The model also suggests that average PPG over three with the evening meal and has a longer half-life, meaning meals is lower in older patients initiated with BIAsp 30 that basal insulin should have advantages over BIAsp 30 in than with basal insulin. This is due to BIAsp 30 containing terms of FPG levels measured before breakfast. a rapid-acting insulin component to specifically target PPG and its administration once or twice daily before meals. The difference between the two treatments was less marked in younger patients, but again, this finding Conclusions should be interpreted with some caution because less Despite the limitations of this analysis, the results of this data from younger individuals were included. The ability study provide useful information to guide treatment of biphasic insulin analogs to achieve lower mean PPG choices when starting insulin. For example, predicted values than basal insulin analogs is supported by a 3,9,13 HbA1c reductions were better following insulin initiation number of previous studies . with BIAsp 30 than basal insulin in older patients with higher bedtime PG (Figure 1). These data also suggested that BIAsp 30 may be a better option in people over 55 Limitations of this pooled analysis years of age, whatever their bedtime PG (Figure 1). As One limitation of this study was that it only included infor- people with diabetes face progressive islet beta-cell failure mation from Novo Nordisk trials as the analysis required over time, it is likely that older patients with higher bed- access to individual patient data that was not available in time BG may have poor beta-cell function and may there- the public domain. This limited access to data meant that fore need both basal and prandial insulin provided by the included trials were fairly heterogeneous in terms of BIAsp 30. In contrast, younger patients with low bedtime size, duration and specific treatment regimens. This neces- PG may have more residual beta-cell function and need, at sitated the use of a strict methodology to minimize poten- least initially, an insulin targeting basal BG only. This tial confounding factors. In addition, only one of the trials potential difference in beta-cell function may also explain in the pooled analysis provided a direct comparison of the finding that there was less difference in PPG with BIAsp 30 and basal insulin9, so this study may have had BIAsp 30 and basal insulin in younger patients compared a disproportionate effect on the study findings. Another with older patients.

For personal use only. potential limitation of this analysis is that a meal- The finding that a basal insulin is better in individuals time þ basal insulin arm was not included in the analysis; with lower BMI and higher post-breakfast PG (Figure 2) is, however, basal and premix insulin were thought to be more on the surface, not consistent with the islet beta-cell dete- common insulin starter choices in T2DM. Different con- rioration concept. However, it is likely that patients with a comitant OGLD therapies were used in the analyzed trials, lower BMI would be more insulin-sensitive than those which could potentially affect efficacy and safety out- with a higher BMI, and would therefore respond well to comes; however, this confounding factor was controlled basal insulin given at bedtime, whereas those patients for in the data analysis. with higher BMI, implying a higher degree of insulin resis- The comparison of BIAsp 30 and basal insulin is also tance, may do better when both basal and postprandial made difficult because of differences between the two glucose is targeted, enabling higher doses of insulin to treatments. In the Raskin et al. 2005 study9, the targeted be administered and providing better overall glucose Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14 FPG was the same for both insulin glargine and BIAsp 30, reductions. and the two groups achieved nearly identical degrees of Despite these findings, it is important to tailor insulin FPG lowering. This reflects the fact that the daily dose of treatment to the individual. Major hypoglycemic episodes ‘extended-acting’ insulin was nearly identical: 0.55units/ were rare with both treatments, but as has been observed in kg of insulin glargine and 0.57units/kg of aspart protamine this and other treat-to-target studies3,9,13, they were more insulin (70% of the total administered 0.82units/kg of pre- common with premix insulins than with basal insulins. mixed BIAsp 30). Therefore, the difference in the inten- Therefore, this could affect therapy choice in patients sity of the treatments was solely due to the postprandial who are unable to recognize and/or treat the symptoms component of BIAsp 30 (0.25units/kg daily). Likewise, the of hypoglycemia. improved HbA1c reduction associated with BIAsp 30 as This study identified patient baseline characteristics compared with insulin glargine in the study can be attrib- that could be useful in determining whether patients uted to the aspart in BIAsp 30 targeting PPG, in addition should be initiated with a basal or premix insulin. The to the FPG targeting provided by the basal component. In results of this study could also help in the planning of summary, BIAsp 30 is designed to provide both basal and future clinical trials by identifying suitable treatment prandial coverage so, by default, will have advantages over groups to target with particular insulin regimens.

! 2010 Informa UK Ltd www.cmrojournal.com Pooled analysis of basal versus premix insulin analog Fonseca et al. 1627 Current Medical Research & Opinion Volume 26, Number 7 July 2010

glucose-lowering drugs in insulin-naive people with type 2 diabetes. Transparency Diabetologia 2008;51:408-16. Declaration of funding 6. Hermansen K, Davies M, Derezinski T, et al. A 26-week, randomized, parallel, Novo Nordisk A/S funded this study. treat-to-target trial comparing insulin detemir with NPH insulin as add-on therapy to oral glucose-lowering drugs in insulin-naive people with type 2 diabetes. Diabetes Care 2006;29:1269-74. Declaration of financial/other relationships 7. Garber A, Wahlen J, Wahl T, et al. for the 1-2-3 study group. Attainment of The authors have disclosed that they and/or the institutions with glycaemic goals in type 2 diabetes with once-, twice-, or thrice-daily dosing which they are associated received funding for educational, with biphasic insulin aspart 70/30 (The 1-2-3 study). Diabetes Obes Metab research and health development activities from the manufac- 2006;8:58-66. turers of premix and basal insulin, including Novo Nordisk. 8. Philis-Tsimikas A, Charpentier G, Clauson P, et al. Comparison of once-daily Some peer reviewers receive honoraria from CMRO for their insulin detemir with NPH insulin added to a regimen of oral antidiabetic drugs review work. The peer reviewers of this paper have disclosed that in poorly controlled type 2 diabetes. Clin Ther 2006;28:1569-81. they have no relevant financial relationships. 9. Raskin P, Allen E, Hollander P, et al. for the INITIATE study group. Initiating insulin therapy in type 2 diabetes: a comparison of biphasic and basal insulin Acknowledgment analogs. Diabetes Care 2005;28:260-5. The authors accept direct responsibility for this paper but are 10. Janka HU, Plewe G, Riddle MC, et al. Comparison of basal insulin added to oral agents versus twice-daily premixed insulin as initial insulin therapy for grateful for the contribution made by Dr Michael Lappin at type 2 diabetes. Diabetes Care 2005;28:254-9. Watermeadow Medical (supported by Novo Nordisk A/S) in 11. Riddle MC, Rosenstock J, Gerich J. Insulin Glargine 4002 Study Investigators. developing a first draft from an agreed outline and in collating The treat-to-target trial: randomized addition of glargine or human NPH insu- comments. lin to oral therapy of type 2 diabetic patients. Diabetes Care 2003;26:3080-6. 12. Robbins DC, Beisswenger PJ, Ceriello A, et al. Mealtime 50/50 basal þ prandial insulin analogue mixture with a basal insulin analogue, both plus

metformin, in the achievement of target HbA1c and pre- and postprandial References blood glucose levels in patients with type 2 diabetes: a multinational, 1. Rodbard HW, Blonde L, Braithwaite SS, et al. AACE Diabetes Mellitus Clinical 24-week, randomized, open-label, parallel-group comparison. Clin Ther Practice Guidelines Task Force. American Association of Clinical 2007;29:2349-64. Endocrinologists medical guidelines for clinical practice for the management 13. Malone JK, Kerr LF, Campaigne BN, et al. Lispro Mixture-Glargine Study of diabetes mellitus. Endocr Pract 2007;13(Suppl. 1):1-68. Group. Combined therapy with insulin lispro Mix 75/25 plus metformin or 2. Nathan DM, Buse JB, Davidson MB, et al. American Diabetes Association; insulin glargine plus metformin: a 16-week, randomized, open-label, cross- European Association for Study of Diabetes. Medical management of hyper- over study in patients with type 2 diabetes beginning insulin therapy. Clin Ther glycemia in type 2 diabetes: a consensus algorithm for the initiation and 2004;26:2034-44. adjustment of therapy: a consensus statement of the American Diabetes 14. Malone JK, Bai S, Campaigne BN, et al. Twice-daily pre-mixed insulin rather

For personal use only. Association and the European Association for the Study of Diabetes. than basal insulin therapy alone results in better overall glycaemic control in Diabetes Care 2009;32:193-203. patients with Type 2 diabetes. Diabet Med 2005;22:374-81. 3. Holman RR, Thorne KI, Farmer AJ, et al. 4-T Study Group. Addition of biphasic, 15. Jacober SJ, Scism-Bacon JL, Zagar AJ. A comparison of intensive mixture prandial, or basal insulin to oral therapy in type 2 diabetes. N Engl J Med therapy with basal insulin therapy in insulin-naı¨ve patients with type 2 2007;357:1716-30. diabetes receiving oral antidiabetes agents. Diabetes Obes Metab 2006;8: 4. Raskin P, Matfin G, Schwartz SL, et al. Addition of biphasic insulin aspart 30 to 448-55. optimized metformin and pioglitazone treatment of type 2 diabetes mellitus: 16. Kazda C, Hu¨lstrunk H, Helsberg K, et al. Prandial insulin substitution The ACTION Study (Achieving Control Through Insulin Plus Oral agents). Diabet with insulin lispro or insulin lispro mid mixture vs. basal therapy with Obes Metab 2009;11:27-32. insulin glargine: a randomized controlled trial in patients with type 2 5. Rosenstock J, Davies M, Home PD, et al. A randomized, 52-week, treat-to- diabetes beginning insulin therapy. J Diabetes Complications 2006;20: target trial comparing insulin detemir with insulin glargine when added to 145-52. Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/22/14

1628 Pooled analysis of basal versus premix insulin analog Fonseca et al. www.cmrojournal.com ! 2010 Informa UK Ltd