Emerging Treatments O R I G I N A L and Technologies A R T I C L E Variability of the Metabolic Effect of Soluble Insulin and the Rapid-Acting Insulin Analog Insulin Aspart LUTZ HEINEMANN, PHD CHRISTIAN WEYER, MD MARTIN RAUHAUS, MD SEBASTIAN HEINRICHS, MD TIM HEISE, MD OBJECTIVE — To study the intra- and interindividual variability of the metabolic activity of soluble insulin and of the rapid-acting insulin analog insulin aspart after subcutaneous injection. RESEARCH DESIGN AND METHODS — A total of nine healthy male volunteers received subcutaneous injections of soluble insulin (0.2 U/kg) in the abdominal region on each of the four study days. Another 10 volunteers received an injection of insulin aspart four times. Glucose infusion rates necessary to neutralize the blood glucose–lowering effect of the administered insulin were registered during euglycemic glucose clamps (blood glucose 5.0 mmol/l; basal intraveneous insulin infusion 0.15 mU kg 1 min 1) over the subsequent 600 min. We investigated the variation in metabolic activity by calculating coefficients of variation (CVs). RESULTS — In comparison to soluble insulin, subcutaneous injections of insulin aspart led to a more rapid onset of action and a shorter duration of action. Subcutaneous injection of the insulin preparations resulted in intraindividual CVs of the summary measures between 10 and 30% (soluble insulin vs. insulin aspart: maximal metabolic activity 15 ± 7 vs. 16 ± 10%, time to maximal metabolic activity 14 ± 10 vs. 11 ± 6%; NS between the preparations [means ± SD]). The decline to half-maximal activity after maximal activity showed a lower intraindividual CV with insulin aspart (19 ± 9 vs. 11 ± 5%; P = 0.018). The interindividual CVs were higher than the intraindividual CVs (26 vs. 28, 23 vs. 19, and 26 vs. 17%). Generally, the pharmacodynamic variability was higher than the pharmacokinetic variability. For the pharmacokinetic measures, the intra- and interindividual variability in tmax was lower for insulin aspart than for soluble insulin. CONCLUSIONS — The metabolic effect of soluble insulin shows an intraindividual variability of 10–20% in healthy volunteers, even under strictly controlled experimental conditions. The overall variability of action of insulin aspart was comparable to that of soluble insulin. Diabetes Care 21:1910–1914, 1998 t is daily clinical experience that subcutaneous injection of insulin leads to a highly variable metabolic effect. This variability hampers the strive for optimal metabolic control. The first quantitative estimation of the variability of insulin I absorption was performed in 1959 (1). This study, as in most other studies performed later on this topic, did not investigate the variability of insulin action, but did investigate the variability of insulin absorption from the subcutaneous depot (2–7). From the Department of Metabolic Diseases and Nutrition, WHO Collaborating Centre for Diabetes, Heinrich-Heine University, Düsseldorf, Germany. Address correspondence and reprint requests to Dr. Lutz Heinemann, Department of Metabolic Diseases and Nutrition, Heinrich-Heine University Düsseldorf, P.O. Box 101007, 40001 Düsseldorf, Germany. E-mail: lutz.heinemann@uni-duesseldorf.de. Received for publication 15 April 1998 and accepted in revised form 17 September 1998. L.H. is a member of an advisory panel for Novo Nordisk. Abbreviations: AUC, area under the curve; CV, coefficient of variation; GIR, glucose infusion rate; GIRmax, maximal glucose infusion rate; Cmax, maximal serum insulin concentration; tmax, time to GIRmax. A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances. 1910 For clinical purposes, quantitative data about the intraindividual variability of the metabolic effect of subcutaneous injected insulin should be helpful. To quantify the intraindividual variability of the metabolic effect, the euglycemic clamp technique was used in two studies (8,9). In the study by Galloway et al. (8), the metabolic effect of long-acting zinc insulin was investigated. Ziel et al. (9) studied soluble insulin, but on two study days and with a study duration of 360 min only. Thus to date, knowledge about the variability of the metabolic effect of currently available short-acting insulin preparations is limited. Subcutaneous injection of rapid-acting insulin analogs like insulin aspart and insulin lispro may lead to a reduced variability of insulin action due to the weaker self-association tendency of insulin monomers into hexamers (10). Insulin aspart was shown to have a more rapid onset of action and a shorter duration of action after subcutaneous injection than soluble insulin (11,12). In clinical experiments, the use of insulin lispro was reported to result in a reduced variability of the metabolic effect (13). Nevertheless, it has not been investigated under controlled experimental conditions, so far, if the variability of the metabolic effect of rapid-acting insulin analogs is lower than that of soluble insulin. We studied the variability of insulin action and of insulin absorption after injecting identical doses of soluble insulin or insulin aspart subcutaneously into the abdominal region of healthy volunteers on four study days. RESEARCH DESIGN AND METHODS Subjects A total of 20 healthy male volunteers (age 26 ± 1 [range 23–28] years; BMI 22.8 ± 1.9 [18.9–26.0] kg/m2) participated in this double-blind study. After receiving detailed oral and written information, all volunteers undersigned an informed consent. The study was performed according to the principles of Good Clinical Practice and the declaration of Helsinki and was approved by the local ethical committee. DIABETES CARE, VOLUME 21, NUMBER 11, NOVEMBER 1998 Heinemann and Associates Study protocol During the morning of each of the four study days, the volunteers were connected to a Biostator after having fasted for at least 12 h. Blood glucose was kept constant at a target level of 5.0 mmol/l by means of the euglycemic glucose clamp technique (14). A low-dose basal intravenous insulin infusion of 0.15 mU kg 1 min 1 was maintained continuously throughout the whole experiment in order to establish comparable baseline insulin levels on all study days. Baseline glucose requirements were registered for 90 min, before insulin was injected subcutaneously by means of a syringe (0.3-ml Microfine IV , Becton Dickinson, Heidelberg, Germany; needle length 13 mm) into the abdominal region. Glucose infusion rates (GIR) necessary to keep blood glucose constant were registered during the subsequent 600 min. The four study days were separated by at least 7 days. The volunteers were instructed to keep their body weight constant and to abstain from strenuous physical exercise during the study period. Identical insulin doses of 0.2 U/kg body weight (14.4 ± 1.6 [range 12–17] U) were injected each time with the same injection technique in the umbilical region by the same investigator (L.H.). The needle was inserted at a 45° angle into a lifted skinfold in order to ensure subcutaneous injection. Nine volunteers received subcutaneous injection of soluble insulin on each of the four study days (Actrapid HM, U100, Novo Nordisk, Bagsvaerd, Denmark), whereas 10 other volunteers received injections of insulin aspart (B28Asp; U100, Novo Nordisk). The 10th volunteer of the soluble insulin group developed an acute otitis media before the first study day, which had to be treated with antibiotics and led to this subject dropping out of the study. The two groups of volunteers did not differ with respect to age and BMI. Blood samples for estimation of plasma glucose, serum insulin, and serum C-peptide concentrations were collected in 30-min intervals before the injection, and thereafter in intervals of 10 min (0–60 min after injection), 15 min (60–120 min), and 30 min (120–600 min), respectively. Serum insulin and serum C-peptide concentrations were estimated by commercial radioimmunoassays (Pharmacia Insulin RIA 100, Pharmacia Uppsala, Sweden, intra-assay coefficient of variation [CV] 4.9% and interassay CV 4.9%) and an enzyme-linked immunosorbent assay (ELISA) (DAKO C-peptide, DAKO Diagnostics, Cambridgeshire, U.K.; intra-assay CV 6.4% and interassay CV 8.7%) by Medi-Lab, Copenhagen, Denmark. The Pharmacia Insulin RIA 100 does not measure the concentration of insulin aspart proportionally to human insulin concentrations. Therefore, the following corrections formula was used to calculate the correct insulin aspart concentrations in the blood: Insulin aspartcorrected F 1,503 1,398 insulin aspart (fraction) insulin aspart (fraction) with F denoting the dilution factor. Statistical methods A lognormal function was fitted to each of the individual time-action profiles registered (15). This function allowed calculation of the following pharmacodynamic summary measures: maximal glucose infusion rate (GIRmax), time to GIRmax (tmax’), time to half-maximal GIR values before and after GIRmax (early t50% and late t50%), and the area under the curve (AUC) of GIR time profiles. The coefficient of variation (CV = SD / –x) was calculated for the summary measures in order to describe the intra- and interindividual variability of insulin action. Intraindividual CVs were calculated from the results obtained with each volunteer on the four study days. Interindividual CVs were calculated from the individual mean values of all volunteers with an identical insulin preparation. An unpaired, two-sided t test was used for statistical comparison of the intraindividual CV of the summary measures obtained with the two insulin preparations. A P value 0.05 was used as significance level. To describe the individual variability of the summary measures graphically, the individual means and their SDs were presented (16). Variability of insulin absorption was studied by fitting a polynomial function to the individual serum insulin concentration profiles. The following pharmacokinetic summary measures were estimated graphically: maximal serum insulin concentration (Cmax), time to Cmax (tmax). The AUCs were calculated for different time periods under the individual serum insulin profiles. The variability of these pharmacokinetic summary measures was studied as described for the pharmacodynamic parameters. The sample size assessment was based on the intrasubject variation comparison. With 20 subjects and four replications per DIABETES CARE, VOLUME 21, NUMBER 11, NOVEMBER 1998 subject, a true ratio of SDs of 0.5 and 2.0 would be discovered with 96% probability at a 5% significance level. RESULTS Blood glucose Blood glucose concentrations were kept constant throughout the glucose clamp experiments at identical levels (soluble insulin vs. insulin aspart 5.0 ± 0.3 vs. 5.0 ± 0.2 mmol/l). Time-action profile Baseline glucose infusion rates were 1.0 ± 0.5 mg kg 1 min 1 with soluble insulin and 0.9 ± 0.5 mg kg 1 min 1 with insulin aspart (NS). As shown previously, the time-action profile of soluble insulin is characterized by a later onset of action, a later maximal effect, and later decline back to baseline values in comparison to the metabolic effects registered after subcutaneous injection of insulin aspart (P 0.01; Fig. 1A, Table 1). The maximal metabolic effect registered was not different after subcutaneous injection of soluble insulin and insulin aspart. Injection of insulin aspart resulted in a higher AUC under the GIR profiles in the first 2 h after injection (P 0.001). The overall metabolic effect over 10 h was comparable (Table 1). The intraindividual variability of the summary measures of soluble insulin and insulin aspart were not significantly different (Table 1), only the decline to half-maximal activity after maximal activity showed a lower intraindividual variability with insulin aspart in comparison to soluble insulin. The interindividual CVs were 5–20% higher than the intraindividual ones (Table 1). The intra- and interindividual variability of some summary measures are given in Fig. 2A–E in a special presentation form. Summary measures do not provide information about the variability of insulin action over time after insulin injection. Calculation of the mean intraindividual CV for each of the two insulin preparations showed that the CV was 25% for both preparations between 60 and 360 min after injection (Fig. 2F). Serum insulin profiles Baseline serum insulin levels were 61 ± 11 pmol/l with soluble insulin and 55 ± 8 pmol/l with insulin aspart (NS). Serum insulin concentrations after subcutaneous injection of insulin aspart increased more rapidly and to higher concentrations in comparison with those obtained with solu1911 Variability of insulin action ble insulin (P 0.001; Fig. 1B; Table 1). The AUCs under the serum insulin profile were higher during the first 2 h after injection of insulin aspart (P 0.001); however, the AUCs were comparable over the whole study duration. With insulin aspart, the intraindividual variability of tmax as well as the interindividual CVs for Cmax, tmax, and AUC0–120 were lower than with soluble insulin. The intra- and interindividual variability for the AUCs were lower for the pharmacokinetic parameters in comparison to the pharmacodynamic summary measures. Serum C-peptide levels were 0.25 ± 0.06 nmol/l after injection of regular insulin and 0.27 ± 0.10 nmol/l with insulin aspart (NS). Endogenous insulin secretion was lowered to 63 ± 20% of basal levels with regular insulin and to 66 ± 29% with insulin aspart (NS). CONCLUSIONS — This is the first investigation looking at the variability of insulin action with a fourfold administration of insulin and an observation period covering the whole duration of insulin action. Repeated subcutaneous injection of soluble insulin in identical doses resulted in an intraindividual variability of the elicited metabolic effect between 10 and 30% and Figure 1—Glucose infusion ratesA) ( and serum insulin concentrations B) (after subcutaneous injecan ininterindividual variability between 20 tion of 0.2 U/kg body weight of the rapid-acting insulin analog, insulin aspart, or of soluble insulin and 50%. Comparison of both insulin healthy male volunteers on four study days (mean ± SEM). Table 1—Intra- and interindividual variability of the metabolic effect and the serum insulin concentrations after subcutaneous injection of soluble insulin or insulin aspart Soluble insulin (n = 9) Intraindividual Interindividual Mean ± SD CV (%) CV (%) Parameter Pharmacodynamic summary measures GIRmax (mg kg 1 min tmax (min) Early t50% (min) Late t50% (min) AUC 0–120 min (mg kg 1 0–600 min (mg kg 1 Pharmacokinetic summary measures Cmax (pmol/l) tmax (min) AUC 0–120 min (nmol l 1 0–600 min (nmol l 1 1) [120 min] 1) [600 min] 1) [120 min] 1) [600 min] 1) Insulin aspart (n = 10) Intraindividual Interindividual Mean ± SD CV (%) CV (%) 9.5 ± 2.3 156 ± 29 61 ± 12 387 ± 68 15 ± 7 14 ± 10 16 ± 10 19 ± 9 26 23 25 26 11.2 ± 2.8 104 ± 16 41 ± 8 264 ± 34 16 ± 10 11 ± 6 16 ± 9 11 ± 5* 28 19† 25† 17† 533 ± 216 3,011 ± 548 27 ± 22 13 ± 3 44 21 861 ± 270 2,840 ± 566 21 ± 12 18 ± 13 36† 25 195 ± 41 129 ± 36 19 ± 8 24 ± 10 28 37 341 ± 40 70 ± 9 14 ± 9 15 ± 6* 18† 20† 16.1 ± 2.9 52.9 ± 5.9 19 ± 8 14 ± 10 31 17 30.8 ± 3.2 52.3 ± 7.2 16 ± 8 15 ± 7 18† 19 The summary measures were estimated after subtraction of baseline glucose infusion rates and serum insulin concentrations. *P 0.001 vs. soluble insulin. 1912 0.05 vs. soluble insulin; †P DIABETES CARE, VOLUME 21, NUMBER 11, N OVEMBER 1998 Heinemann and Associates Figure 2—To describe the intraindividual and the interindividual variability seen with different summary measures A–E), the ( SD of the four measur ements with each volunteer in relation to the mean value are plotted. The SD of the mean value of all individual mean (— values ) describe the interindiIn the mean intraindividual coefficient of variation vidual variability, whereas the SD of all the individual SDs ( ) describe the intraindividual variability. F, of the glucose infusion rates over time with the two insulin preparation are given. preparations studied revealed that the variability of action of insulin aspart was comparable to that of soluble insulin for most summary measures. There are two possible explanations for the variability of insulin action: 1) pharmacokinetic variability (i.e., same doses of insulin lead to different plasma insulin concentrations), and 2) pharmacodynamic variability (i.e., similar plasma insulin concentrations induce different metabolic effects). Concerning the variability of insulin absorption (the pharmacokinetic part), a number of studies have been performed using radioactively labeled soluble insulin (1–7). With this semiquantitative method, an intraindividual CV of 15% and an interindividual CV of 30% has been estimated for the time to decline to half-maximal radioactivity over the insulin depot in the skin (T50%) for soluble insulin (7). Direct measurement of the serum insulin concentrations after subcutaneous insulin injection revealed a considerably higher intraindividual variability of Cmax and tmax of 64 and 107% (soluble insulin), 28 and 34% DIABETES CARE, VOLUME 21, NUMBER 11, NOVEMBER 1998 (lente insulin), and 44 and 68% (NPHinsulin) (17). Insulin absorption from the subcutaneous depot might also vary from injection site to injection site caused by the anatomical properties (e.g., local blood flow) and/or the local degradation of insulin (6,18,19). Injection of human insulin into the abdominal region resulted in a lower variability than that observed after injection of the respective porcine insulin preparation in the deltoid region (8). The variability of insulin action (the pharmacodynamic part) was studied quan1913 Variability of insulin action titatively with the euglycemic glucose clamp technique in two studies only (8,9). Galloway et al. (8) found intra- and interindividual CVs between 35 and 55% for the calculated summary measures for slow-acting insulin preparations employing an unbalanced study design with eight healthy volunteers. Ziel et al. (9) studied the variability of insulin absorption and insulin action with soluble insulin employing a similar study protocol as in our study, but with two study days and a study duration of 360 min only. In this study, the intraindividual CV for the total AUC below the insulin profile was 11%, which was lower than the CV for the AUC below the time-action profile (23%). Also in our study, the variability of the pharmacokinetic summary measures was lower than that of the pharmacodynamic parameters. It is reasonable to assume, that in daily life the variability of insulin action in diabetic patients is even higher than that observed in healthy volunteers studied under controlled experimental conditions. Factors like insulin antibodies, changes in metabolic control, or previous hypoglycemic episodes with their impact on the actual insulin sensitivity were ruled out by our study design. The high variability of the metabolic response to application of identical insulin doses is a serious problem for insulin-treated diabetic patients and hamper the achievement of a good metabolic control without suffering from hypoglycemic events. The poor reproducibility of the metabolic effects of subcutaneous insulin therapy might be a source of frustration and uncertainty in many patients. Therefore, development of insulin preparations with a lower variability of action remains an important task for the future. From a clinical study with insulin lispro it was anticipated that the more rapid absorption of the insulin analog might result in a lower variability of insulin absorption and probably also insulin action in comparison to soluble insulin (13). Nevertheless, we did not observe a lower variability of insulin action after subcutaneous injection 1914 of insulin aspart under controlled experimental conditions, with the exception of the decline of the metabolic effect induced. In summary, under experimental conditions in healthy volunteers, subcutaneous injection of soluble insulin resulted in an intra- and interindividual variability in insulin action of 30%. The variability of action of insulin aspart was equivalent to that of soluble insulin. 9. 10. Acknowledgments — We gratefully acknowledge the skillful help of C. Gottschalk, A. Brodeßer, and M. Schreier during this study. We thank Dr. R. Bender for fitting the lognormal functions to the GIR curves and for the calculation of the summary measures. We thank Novo Nordisk, Mainz, Germany, and Copenhagen, Denmark, for their support during the study. References 1. Moore EW, Mitchell ML, Chalmers TC: Variability in absorption of insulin-I-131 in normal and diabetic subjects after subcutaneous and intramuscular injection. J Clin Invest38:1222–1227, 1959 2. Binder C: Absorption of injected insulin: a clinical-pharmacological study. Acta Phar macol Toxicol(Copenhagen) 2:1–84, 1969 3. Lauritzen T, Faber OK, Binder C: Variation in 125I-insulin absorption and blood glucose concentration. Diabetologia17:291–295, 1979 4. Lauritzen T, Binder C, Faber OK: Importance of insulin absorption, subcutaneous blood flow, and residual beta-cell function in insulin therapy. Acta Paediatr Scand283 (Suppl.):81–85, 1980 5. Kolendorf K, Bojsen J, Deckert T: Clinical factors influencing the absorption of 125INPH insulin in diabetic patients. Horm Metab Res15:274–278, 1983 6. Binder C, Lauritzen T, Faber O, Pramming S: Insulin pharmacokinetics. Diabetes Care 7:188–199, 1984 7. de Meijer PH, Lutterman JA, van Lier HJ, van’t Laar A: The variability of the absorption of subcutaneously injected insulin: effect of injection technique and relation with brittleness. Diabet Med7:499–505, 1990 8. Galloway JA, Spradlin CT, Howey DC, Dupre J: Intrasubject differences in phar- 11. 12. 13. 14. 15. 16. 17. 18. 19. macokinetic and pharmacodynamic responses: the immutable problem of present-day treatment. In Diabetes1985. Serrano-Rios M, Lefebvre PJ, Eds. New York, Excerpta Medica, 1986, p. 877–886 Ziel FH, Davidson MB, Harris MD, Rosenberg CS: The variability in the action of unmodified insulin is more dependent on changes in tissue insulin sensitivity than on insulin absorption. Diabet Med5:662–666, 1988 Brange J, Owens DR, Kang S, Volund A: Monomeric insulins and their experimental and clinical implications. Diabetes Care 13:923–954, 1990 Heinemann L, Heise T, Jorgensen LN, Starke AAR: Action profile of the rapid-acting insulin analogue B28Asp. Diabet Med 10:535–539, 1993 Heinemann L, Kapitza C, Starke AAR, Heise T: Time-action profile of the insulin analogue B28Asp. Diabet Med13:683–684, 1996 Antsiferov M, Woodworth JR, Mayorov A, Ristic S, Dedov II: Lower within patient variability in postprandial glucose excursion with regular insulin (Abstract). Dia betologia38 (Suppl. 1):A190, 1995 Starke AAR, Heinemann L, Hohmann A, Berger M: The action profiles of human NPH insulin preparations. Diabet Med 6:239–244, 1989 Bender R, Heinemann L: Fitting non-linear regression models with correlated errors to individual pharmacodynamic data using SAS software. J Pharmacokinet Biopharm 23:87–100, 1995 Bland JM, Altman DG: Measurement error. BMJ 312:1654, 1996 Galloway JA, Spradlin CT, Nelson RL, Wentworth SM, Davidson JA, Swarner JL: Factors influencing the absorption, serum insulin concentration, and blood glucose responses after injections of regular insulin and various insulin mixtures. Diabetes Care 4:366–376, 1981 Hildebrandt P: Subcutaneous absorption of insulin in insulin-dependent diabetic patients: influence of species, physico-chemical properties of insulin and physiological factors. Dan Med Bull38:337–346, 1991 Linde B: The pharmacokinetics of insulin. In Textbook of Diabetes . Pickup JC, Williams G, Eds. London, Blackwell Scientific, 1991, p. 371–383 DIABETES CARE, VOLUME 21, NUMBER 11, N OVEMBER 1998