DIABETES TECHNOLOGY & THERAPEUTICS Volume 10, Number?, 2008 Mary Ann Liebert, Inc. DOI: 10.1089/dia.2007.0297 LIND ET AL. GLYCEMIC EFFECTS OF INSULIN LISPRO The Effect of Insulin Lispro on Glycemic Control in a Large Patient Cohort Marcus Lind, M.D., 1 Martin Fahlén, M.D., Ph.D., 2 Michael Happich, Ph.D., 3 Anders Odén, Ph.D., 4 and Björn Eliasson, M.D., Ph.D. 5 1 Department of Medicine, Uddevalla Hospital, Uddevalla, Sweden. 2 Department of Medicine, Kungälv Hospital, Kungälv, Sweden. 3 Health Outcomes, Lilly Deutschland GmbH, Bad Homburg, Germany. 4 Department of Mathematical Sciences, Chalmers University of Technology, Göteborg, Sweden. 5 Department of Internal Medicine, Sahlgrenska University Hospital, Göteborg, Sweden.
ABSTRACT Background: The use of rapid-acting insulin analogues and regular insulin differs considerably in countries throughout the world. We therefore studied how glycemic control has been affected by using insulin lispro in clinical practice over 5 years in 14 hospitals in Sweden. Methods: We used a time period when most patients had not changed the basal insulin, but only the mealtime insulin. Accordingly the most recent years were not suitable since many patients had changed basal insulin from NPH to glargine or determir. We therefore analyzed the metabolic consequences on glycosylated hemoglobin (HbA1c) when changing from regular insulin to insulin lispro from 1997 and during the following 5 years. We studied 1,069 patients with diabetes taking NPH insulin as basal insulin and at least three daily injections of regular insulin, of whom 423 changed their mealtime insulin to insulin lispro and 646 controls who continued with regular insulin. Results: Patients changing to insulin lispro on average decreased by 0.19% units more in HbA1c than those remaining on regular insulin. The effect was most pronounced in patients with high HbA1c even after controlling for regression to the mean. A beneficial effect of insulin lispro was also indicated since patients had the same level of HbA1c during a long period of time with regular insulin but then dropped when changing to insulin lispro. Conclusions: This study indicates that insulin lispro has had a beneficial and persisting effect on glycemic control when used in patients with diabetes on multiple daily injections of insulin in clinical practice. 2
INTRODUCTION C LINICAL PRACTICE RECOMMENDATIONS assert the importance of good glycemic control in diabetes mellitus. 1 The different available mealtime insulins have different hypoglycemic effects, providing tools to tailor the treatment of type 1 and type 2 diabetes patients. 2,3 However, the clinical value of rapidacting insulin analogues versus regular insulin has been critically evaluated with a slight beneficial effect of rapid-acting insulin analogues on glycosylated hemoglobin (HbA1c) and hypoglycemia. 4 The use of rapid-acting insulin analogues and regular insulin differs considerably in countries throughout the world. A quality control of the possible beneficial effects of rapid-acting insulin analogues in large patient groups over a long time is lacking. Since patients with type 1 diabetes and many patients with type 2 diabetes use mealtime insulins every day, there is a need for a study from clinical practice to control for how the glycemic control has been affected in clinical practice. This would also indicate what possible beneficial effects there have been on complications of diabetes since rapid-acting insulins were introduced into the daily care of diabetes patients. The aim of this observational study was thus to study the effects of insulin lispro versus regular insulin on glycemic control in clinical practice. We wanted to use a time period when most patients had not changed the basal insulin, but only the mealtime insulin, so that only one medication was changed. Accordingly the most recent years were not suitable since many patients had changed basal insulin from NPH insulin to insulin glargine or determir. We therefore analyzed the metabolic consequences on HbA1c when changing from regular insulin to insulin lispro from 1997 and during the following 5 years. We studied 1,069 diabetic patients with NPH insulin as basal insulin and at least three daily injections of regular insulin of whom 423 changed their mealtime insulin to insulin lispro and 646 controls who continued with regular insulin. Subjects and Methods Diab-Base The study was approved by the ethics committee of Gothenburg University (Gothenburg, Sweden). From a medical records system and quality improving system called Diab-Base we collected extensive patient data from 17 diabetes clinics from 4 15 years backwards from year 2007. In total, patient data of 20,000 type 1 and type 2 diabetes patients were collected in 2007. From this system it is possible to track different treatments, doses of medications, and the date when treatments have been changed for individual patients. 5 The system also provides extensive information on the HbA1c levels measured in clinical practice, grade of microalbuminuria, blood pressure, blood lipids, and other common risk variables and laboratory markers of diabetes. Information 3
on retinopathy, neuropathy, nephropathy, foot ulcers, amputations, myocardial infarction, stroke, and autonomic complications can also be tracked in the system. Evaluation of observational data When evaluating HbA1c values one source of error could be change in laboratory method or imprecision, disturbance, or systematic error in such a method. 6 This could be particularly difficult to control for in an observational study since measures can then be made at different hospitals, at different dates, and for different treatments. One possible way of diminishing these risks is the evaluation of HbA1c at the same point of time for all patients included and for the treatments compared. Otherwise, there could be a risk that one treatment evaluated at a certain point of time will be coupled with a lower HbA1c because of occasional disturbances in the laboratory assay. Another way of controlling for such a problem could be to follow the patients longitudinally over a long period of time and see that HbA1c is stable before and after the change to a certain type of therapy. Another important factor is possibly that patients are on stable treatment at baseline, that is, they have not recently been put on insulin, thereby not being optimally titrated and the insulin sensitivity not stabilized. The baseline HbA1c could then be falsely high or low for the actual treatment since the patients have not had time to be titrated or if recently diagnosed the insulin need might be lower than usual. This risk can possibly be minimized by the patient having been on the actual treatment for a long period of time. After the change of treatment the time for measuring of HbA1c must be at least a couple of months later so that HbA1c could have had a possibility to change. Inclusion Since we wanted to study the effects of insulin lispro and regular insulin we tracked the database system during those years when basal insulin in most patients was unchanged but mealtime insulins were often changed from regular insulin to rapid-acting insulin analogues. Insulin lispro was introduced into the Swedish market in 1997, and during the following years most patients used NPH insulin as basal insulin. In 1996 there were 14 diabetes clinics in Sweden using the Diab-Base management system. 6 In total, there were patient records of 5,920 patients with diabetes at this point of time, and for inclusion patients should remain in the system over 5 years and to avoid possible bias in evaluation of HbA1c: 4
1. be on multiple daily injections of insulin (MDI) with regular insulin on January 1, 1997 2. be on NPH as basal insulin on January 1, 1997 and remain on NPH insulin until the end of 2001 3. remain on regular insulin or change to insulin lispro during the study period 4. on January 1, 1997 have been on insulin treatment for more than 300 days 5. at least one HbA1c measurement had to exist at the same points of time for all patients, that is, at baseline in 1996 and in 2001 6. the HbA1c value of patients changing to insulin lispro in 2001 had to be taken at least 2 months after the change to insulin lispro Patients changing to any other type of treatment such as continuous subcutaneous insulin infusion, insulin glargine, or insulin aspart were excluded. Design Thus two groups were studied: (1) patients continuing with the NPH and the regular insulin between 1996 and 2001 and (2) patients continuing with the NPH insulin, but changing mealtime insulin from regular insulin to insulin lispro in the period 1996 2001. Several HbA1c values existed during the study period for most patients, making it possible to follow these values before and after the change from regular insulin to insulin lispro. The statistical analysis controlled for several factors. It made it possible to compare insulin lispro with regular insulin for patients with the same level of HbA1c at baseline. It also controlled for differences in age, body mass index (BMI), sex, duration of insulin treatment, the year when the change of insulin regime was made, type of diabetes, smoking, weight, daily insulin requirement, and number of insulin doses. Diabetes outpatient clinics in Sweden use methods regularly calibrated to the high-performance liquid chromatography Mono-S method ((Mono-S; General Electric Health Care, Uppsala Sweden). HbA1C values can be converted to the Diabetes Control and Complications Trial (DCCT) standard levels using the formula: A1C (DCCT) = 0.923 A1C (Mono-S) + 1.345 (R 2 = 0.998). 7 All HbA1c values presented in this study have been converted to DCCT standard values. The HbA1c standardization in Sweden is handled with the help of Equalis, an organization that regularly tests the validity of the HbA1c methods. Statistical methods 5
Variables are presented as mean ± 1 standard deviation (SD). Paired t test was used for comparisons of HbA1c levels before and after the change to insulin lispro and for those patients remaining on regular insulin. Fisher s exact test and t test were used for comparisons between groups. Multiple regression analysis with HbA1c change as the dependent variable was used for the whole cohort. Among the independent variables the zero-one variable indicating change to lispro was included together with interactions with that variable. The model allowed a comparison of changes in HbA1c for those remaining on regular insulin versus those changing to insulin lispro taking possible confounders into account. All P values were two-tailed. A value of P < 0.05 was considered statistically significant. Results Patient characteristics of the cohort are described in Table 1. Although there were differences between the groups, it was only BMI and HbA1c at baseline that influenced HbA1c levels during the study period. Age, sex, duration of insulin treatment, the year when the change of insulin regime was made, type of diabetes, smoking, weight, daily insulin requirement, and number of insulin doses did not significantly affect HbA1c and were hence not adjusted for in the model. There was a statistical interaction between HbA1c at baseline and change to insulin lispro indicating greater benefits of lispro at higher HbA1c values (Fig. 1). HbA1c levels were stable until the change to insulin lispro but then dropped (Table 2). Patients changing to insulin lispro decreased on average 0.19% units more in HbA1c than those remaining on regular insulin. This effect was estimated from an HbA1c level of 8.14%, and statistical adjustments were made to compare both groups for that HbA1c level and adjust for differences in BMI. This means that in the statistical model patients remaining on regular insulin were stable at 8.14% during the study period, whereas those changing to insulin lispro decreased from 8.14% to 7.95%. Patients changing to insulin lispro increased the number of daily insulin injections significantly (Table 2). We therefore also included change in number of daily insulin injections in the model, but it was not a significant predictor of change in HbA1c in the cohort. Discussion The rapid-acting insulin analogues insulin lispro and insulin aspart have been shown to improve glycemic control in patients treated with MDI and NPH insulin as bedtime insulin in randomized clinical trials. 4 The positive effect was most evident in type 1 diabetes patients, in clinical trials with a longer intervention time 6
than 3 months, and in trials published after 2000. 4 Because of a relatively modest mean effect on glycemic control and only sparse effects on rates of hypoglycemia and quality of life but a higher cost, the role of these insulin analogues has been questioned. 4 It is therefore of interest to investigate whether a metabolic improvement can be observed in daily practice, as has previously been shown for insulin glargine. 8 We followed a historical cohort of 1,069 patients with NPH insulin during MDI treatment with regular insulin over 5 years, of whom 423 changed to insulin lispro during this period of time. The clinicians generally chose to change therapy to insulin lispro in patients with somewhat worse glycemic control, indicated by an HbA1c of 0.15% units higher for these patients. This fact could possibly be explained by patients with a higher level of HbA1c being more motivated to try a new therapy as well as the clinician seeing greater benefits of trying a new treatment in a patient with worse glycemic control. The number of injections increased significantly for patients changing to insulin lispro. The probable explanation of more injections is that the introduction of lispro unmasks a shortage of basal insulin that was earlier covered by regular insulin. During the follow-up period those changing to insulin lispro as mealtime insulin decreased on average by 0.19% units in HbA1c in comparison to those remaining on regular insulin (Fig. 1). This mean effect is comparable with results of randomized clinical trials, 4 and thus this study of data from everyday clinical practice supports a true effect of these treatments, providing valuable tools when individualizing the treatment of patients with diabetes. The beneficial effect of insulin lispro was also supported by the fact that the HbA1c was stable before the change to insulin lispro but then dropped significantly after the change. A difference of 0.19% units of HbA1c at the HbA1c level of 8.14% might at a glance seem small. However, to put it in a national perspective of Sweden with roughly 9 million inhabitants and a diabetes prevalence of around 4%, roughly 1,000 fewer patients would suffer progression in retinopathy each year if we assume that all patients would decrease from an HbA1c level of 8.14% to 7.95%. 9 Knowing that the risk gradients of HbA1c are equally steep for amputations, neuropathy, and nephropathy as for retinopathy and that HbA1c also probably is of importance for myocardial infarction and stroke, one can understand that in a world perspective this effect is of clinical importance. 10 However, it is not possible to cover in this paper whether this decrease in HbA1c and possible reduction in complications of diabetes can be considered cost-beneficial. If an insulin regimen in a patient does not lead to the desired effects such as improved glycemia, ease of use, and less hypoglycemia, the treatment should generally be revised. In observational studies of pharmacological agents there is always a risk of selection bias as well as confounding by indication. 11 To diminish the risk of bias we made several adjustments in the design of the study. 7
We observed that the glycemic level was stable during a long period of time before and after the change to insulin lispro to avoid regression to the mean. Comparisons of HbA1c levels with those remaining on regular insulin were made after considering several potential confounders. Another adjustment was that only patients who were on insulin treatment for at least 300 days were included to avoid unstable HbA1c at the time of diagnosis. We carefully excluded all patients changing to any other form of treatment during the 5 years observed and considered any impact on HbA1c of time when patients changed to insulin lispro. All patients included should have at least one HbA1c measurement before the follow-up period and one in the end so that HbA1c values at the same points of time existed to further control for possible differences in laboratory methods over time. We have used a medical records system especially adapted to diabetes clinics for this study. Laboratory data such as HbA1c and data about medication are of high quality, since HbA1c is regularly measured according to certain routines in clinical practice, and the prescriptions of medication are recorded via the system. Clinical experience, as well as the results of one meta-analysis, with type 1 diabetes patients shows that the risk of hypoglycemia decreases when starting treatment with insulin lispro. 12 The system includes information of hypoglycemia codes that can be used with phrases about hypoglycemia, i.e., codes with the number of difficult hypoglycemias per year and codes with time specifications about insulin coma and other types of severe hypoglycemia. The system reminds the user of the necessity of recording but does not force him or her to do so. This has led to an underreporting, and fewer than half of the patients during the study period had information in the system on this issue, and hypoglycemia was hence not included in the analyses. Gradually we have found that the data quality improves when the system is more mandatory. If you undertake a prospective study, the system can be modified in order to further improve the quality of the data. In this study patients with high HbA1c decreased most in HbA1c after changing treatment to insulin lispro, even after controlling for regression to the mean (Fig. 1). Possibly patients with higher HbA1c had worse compliance to insulin treatment but achieved better compliance with rapid-acting insulin analogues since these were taken in close connection to their meals. Another finding in the prediction analysis was that low BMI led to greater changes in HbA1c for the whole cohort. We have recently described a similar phenomenon for insulin glargine, and possibly it is due to lower insulin resistance and could indicate a general need of careful optimization of insulin titration in those with high BMI. 8 We have also seen that after the introduction of the new basal insulins and the more frequent use of pumps there was a further decrease of HbA1c. 5,8 8
In conclusion, this study indicates that insulin lispro generally has a beneficial effect in clinical care in comparison with regular insulin. It thus supports the positive mean effects observed in meta-analyses. Patients with high HbA1c were those benefiting most from the change from regular insulin to insulin lispro in this patient cohort from Swedish clinical care. This observational study adds support to the concept of individualization of MDI in patients with diabetes taking rapid-acting insulin analogues. Acknowledgments We would like to thank all the nurses, physicians, and other members of the staff who participated in providing data. This study was supported by an unrestricted research grant from Lilly Deutschland GmbH, Germany. References 1. Standards of medical care in diabetes 2006. Diabetes Care 2006;29(Suppl 1):S4 S42. 2. DeWitt DE, Hirsch IB: Outpatient insulin therapy in type 1 and type 2 diabetes mellitus: scientific review. JAMA 2003;289:2254 2264. 3. Hirsch IB: Insulin analogues. N Engl J Med 2005;352:174 183. 4. Plank J, Siebenhofer A, Berghold A, Jeitler K, Horvath K, Mrak P, Pieber TR: Systematic review and meta-analysis of short-acting insulin analogues in patients with diabetes mellitus. Arch Intern Med 2005;165:1337 1344. 5. Fahlén M, Eliasson B, Odén A: Optimization of basal insulin delivery in Type 1 diabetes: a retrospective study on the use of continuous subcutaneous insulin-infusion and insulin glargine. Diabet Med 2005;22:382 386. 6. Hanås R: Psychological impact of changing the scale of reported HbA 1c results affects metabolic control. Diabetes Care 2002;25:2110 2111. 7. Hoelzel W, Weykamp C, Jeppsson JO, Miedema K, Barr JR, Goodall I, Hoshino T, John WG, Kobold U, Little R, Mosca A, Mauri P, Paroni R, Susanto F, Takei I, Thienpont L, Umemoto M, Wiedmeyer HM; IFCC Working Group on HbA1c Standardization: IFCC reference system for measurement of hemoglobin A1c in human blood and the national standardization schemes in the United States, Japan, and Sweden: a method-comparison study. Clin Chem 2004;50:166 174. 9
8. Österbrand M, Fahlén M, Odén A, Eliasson B: A method to predict the metabolic effects of changes in insulin treatment in subgroups of a large population based patient cohort. Eur J Epidemiol 2007;22:151 157. 9. Diabetes Control and Complications Trial Study Group: The relationship of glycemic exposure (HbA1c) to the risk of development and progression of retinopathy in the Diabetes Control and Complications Trial. Diabetes 1995;44:968 983. 10. Stratton IM, Adler AI, Neil HA, Matthews DR, Manley SE, Cull CA, Hadden D, Turner RC, Holman RR: Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ 2000;321:405 412. 11. Signorello LB, McLaughlin JK, Lipworth L, Friis S, Sorensen HT, Blot WJ: Confounding by indication in epidemiologic studies of commonly used analgesics. Am J Ther 2002;9:199 205. 12. Brunelle BL, Llewelyn J, Anderson JH Jr, Gale EA, Koivisto VA: Meta-analysis of the effect of insulin lispro on severe hypoglycemia in patients with type 1 diabetes. Diabetes Care 1998;21:1726 1731. Address reprint requests to: Marcus Lind, M.D. Department of Medicine Uddevalla Hospital SE-451 80 Uddevalla, Sweden E-mail: lind.marcus@telia.com 10
TABLE 1. BASELINE CHARACTERISTICS OF THE PATIENT COHORT Insulin Regular Lispro P value Number of patients 646 423 Age (years) 49.3 (14.5) 40.9 (12.7) P < 0.001 Female (%) 38.60% 49.40% P < 0.001 Type 1 (%) 72.60% 89.90% P < 0.001 BMI (kg/m 2 ) 25.8 (3.8) 25.7 (3.9) NS Duration of diabetes (years) 20.7 (12.8) 17.5 (11.5) P < 0.001 HbA1c (%) 7.99 (1.33) 8.14 (1.36) P < 0.001 Insulin dose (E/kg) 0.67 (0.29) 0.68 (0.24) NS Number of doses 4.06 (0.43) 4.24 (0.63) P < 0.001 Follow-up time (days) 1,794 (103) 1,813 (110) P < 0.01 Days with insulin lispro 1,130 (range 61 1,868) Patients are divided into those continuing with regular insulin during the whole 5-year observational period and those changing treatment to insulin lispro. NS, not significant. 11
TABLE 2. CHANGES IN HBA1C, INSULIN DOSE, AND NUMBER OF DAILY INSULIN INJECTIONS DURING THE STUDY PERIOD Insulin Regular Lispro HbA1c (%) in 1996 7.99 (1.33) 8.14 (1.36) In 1996 At change a 8.19 (1.32) P value (HbA1c at change vs. HbA1c in 1996) NS In 2001 7.98 (1.30) 7.97 (1.21) P value (HbA1c in 2001 vs. 1996) NS P < 0.0001 Insulin dose (E/kg) In 1996 0.67 (0.29) 0.68 (0.24) In 2001 0.73 (0.35) 0.71 (0.24) P value (E/kg in 2001 vs. 1996) P < 0.0001 P < 0.05 Daily injections (number/day) In 1996 4.06 (0.43) 4.24 (0.63) In 2001 4.03 (0.68) 4.42 (0.88) P value (injections in 1996 vs. 2001) NS P < 0.0001 Data are shown for patients continuing with regular insulin over 5 years versus those changing to insulin lispro during the same period of time. All comparisons are within group comparisons with paired t test. NS, not significant. a HbA1c at the time when patients changed from regular insulin to insulin lispro. 12
FIG. 1. Patients with high HbA1c levels had the greatest beneficial effect on HbA1c when changing mealtime insulin from regular insulin to insulin lispro. The model is most valid for HbA1c values between 7% and 9% and must be interpreted with caution outside these values. 13