Insulin Pump Programming: Preventing Dawn Phenomenon in T1D
Healthy volunteers with type 1 diabetes aged 18 to 70 years were recruited from the community to participate in this 8-month longitudinal study. Eligible participants were insulin-dependent for at least 1 year prior to enrollment with a current hemoglobin A1c between 6.5 and 9.0%. Exclusion criteria included pregnancy, renal insufficiency, or the presence of any significant mental or physical disabilities which might affect compliance with the protocol. Participants were also excluded for use of morning-only dosing of basal insulin to avoid introducing the effects of early morning waning insulin levels on CGM data. For the purposes of this investigation, the recruited population was composed of 2 groups of subjects: (1) type 1 diabetic subjects treated with multiple daily injections (MDI) of insulin (n = 12), and (2) type 1 diabetic subjects treated with CSII (n = 28). The latter group of subjects was stratified by the presence or absence of programming for an early morning increase in insulin delivery via the insulin pump (CSII programmers [n = 20] versus CSII nonprogrammers [n = 8]) occurring between 3:00 and 7:00 AM. This study was approved by the University of New Mexico Health Sciences Center Human Research Review Committee, and all subjects provided written informed consent prior to entering the study.
All participants in the current study presented for 8 monthly visits and had a CGM (iPro™, Medtronic, Minneapolis, MN) placed at visits 2, 4, 6, and 8, with instructions for CGM removal provided during the initial visit. Participants were blinded to the CGM data, although frequent self-monitoring of blood glucose levels was required. After 5 days of monitoring, the CGM was removed and delivered to the investigators for analysis. All CGM data were read independently by 2 investigators who were blinded to subject categorization in order to minimize bias in analysis.
Hemoglobin A1c was assayed at the University of New Mexico General Clinical Research Center utilizing a DCA 2000 (Bayer HealthCare LLC, Elkhart, IN) at each monthly visit during the 8-month study.
The design of the current study permitted determination of: (1) the reproducibility of the dawn phenomenon in type 1 diabetic subjects, and (2) the effects of programming insulin pumps for an early morning increase in insulin delivery on the occurrence of the dawn phenomenon and hypoglycemia. In order to determine the reproducibility of the dawn phenomenon, we analyzed CGM data from the MDI and CSII nonprogrammer groups. CSII programmers were not included in this determination to avoid potential confounding effects of early morning increases in insulin delivery on the occurrence of the dawn phenomenon. To investigate the effects of programming an early morning increase in insulin delivery on the occurrence of the dawn phenomenon and hypoglycemia, we compared the CSII programmers to nonprogrammers. The dawn phenomenon was defined as an early morning increase in CGM-determined glucose between the overnight nadir (occurring between 2:00 and 5:00 AM) and the breakfast meal (occurring between 5:00 and 10:00 AM) of 10 mg/dL, as previously reported. Early morning increases in glucose of greater than 10 mg/dL preceded by hypoglycemia were excluded from analysis to avoid introducing the confounding effect of counterregulatory responses to hypoglycemia in our data interpretation. These occurrences were, however, recorded for determination of hypoglycemic rates. Hypoglycemia was defined as a CGM-determined glucose of less than 70 mg/dL occurring at any time between midnight and the breakfast meal. Subjects wore the CGM for a maximum of 5 days, permitting analysis of up to 4 overnight periods. Overnight observations that were incomplete due to unintended hardware interruption or compromised by overnight eating or bolus insulin administration were excluded from analysis. Additional outcome measures included hemoglobin A1c, postprandial glucose excursions following the breakfast meal (2-hour postprandial glucose – preprandial glucose), and 24-hour average glucose levels as determined by CGM.
For each of the variables of interest, there were up to 20 observations for each patient. Summary measures for each patient were computed as variable means for continuous variables and percent occurrence for binary variables. Both types of summary measures were treated as continuous. CSII programmers were compared to nonprogrammers on the summary measures using 2-sample t tests with adjustment for heterogeneity of variance. Pairwise Pearson correlations were computed. Statistical significance was taken as a 2-sided P value <.05. Calculations were performed in SAS ver. 9.3.
Methods
Study Cohort
Healthy volunteers with type 1 diabetes aged 18 to 70 years were recruited from the community to participate in this 8-month longitudinal study. Eligible participants were insulin-dependent for at least 1 year prior to enrollment with a current hemoglobin A1c between 6.5 and 9.0%. Exclusion criteria included pregnancy, renal insufficiency, or the presence of any significant mental or physical disabilities which might affect compliance with the protocol. Participants were also excluded for use of morning-only dosing of basal insulin to avoid introducing the effects of early morning waning insulin levels on CGM data. For the purposes of this investigation, the recruited population was composed of 2 groups of subjects: (1) type 1 diabetic subjects treated with multiple daily injections (MDI) of insulin (n = 12), and (2) type 1 diabetic subjects treated with CSII (n = 28). The latter group of subjects was stratified by the presence or absence of programming for an early morning increase in insulin delivery via the insulin pump (CSII programmers [n = 20] versus CSII nonprogrammers [n = 8]) occurring between 3:00 and 7:00 AM. This study was approved by the University of New Mexico Health Sciences Center Human Research Review Committee, and all subjects provided written informed consent prior to entering the study.
Continuous Glucose Monitoring
All participants in the current study presented for 8 monthly visits and had a CGM (iPro™, Medtronic, Minneapolis, MN) placed at visits 2, 4, 6, and 8, with instructions for CGM removal provided during the initial visit. Participants were blinded to the CGM data, although frequent self-monitoring of blood glucose levels was required. After 5 days of monitoring, the CGM was removed and delivered to the investigators for analysis. All CGM data were read independently by 2 investigators who were blinded to subject categorization in order to minimize bias in analysis.
Laboratory Analysis
Hemoglobin A1c was assayed at the University of New Mexico General Clinical Research Center utilizing a DCA 2000 (Bayer HealthCare LLC, Elkhart, IN) at each monthly visit during the 8-month study.
Study Design
The design of the current study permitted determination of: (1) the reproducibility of the dawn phenomenon in type 1 diabetic subjects, and (2) the effects of programming insulin pumps for an early morning increase in insulin delivery on the occurrence of the dawn phenomenon and hypoglycemia. In order to determine the reproducibility of the dawn phenomenon, we analyzed CGM data from the MDI and CSII nonprogrammer groups. CSII programmers were not included in this determination to avoid potential confounding effects of early morning increases in insulin delivery on the occurrence of the dawn phenomenon. To investigate the effects of programming an early morning increase in insulin delivery on the occurrence of the dawn phenomenon and hypoglycemia, we compared the CSII programmers to nonprogrammers. The dawn phenomenon was defined as an early morning increase in CGM-determined glucose between the overnight nadir (occurring between 2:00 and 5:00 AM) and the breakfast meal (occurring between 5:00 and 10:00 AM) of 10 mg/dL, as previously reported. Early morning increases in glucose of greater than 10 mg/dL preceded by hypoglycemia were excluded from analysis to avoid introducing the confounding effect of counterregulatory responses to hypoglycemia in our data interpretation. These occurrences were, however, recorded for determination of hypoglycemic rates. Hypoglycemia was defined as a CGM-determined glucose of less than 70 mg/dL occurring at any time between midnight and the breakfast meal. Subjects wore the CGM for a maximum of 5 days, permitting analysis of up to 4 overnight periods. Overnight observations that were incomplete due to unintended hardware interruption or compromised by overnight eating or bolus insulin administration were excluded from analysis. Additional outcome measures included hemoglobin A1c, postprandial glucose excursions following the breakfast meal (2-hour postprandial glucose – preprandial glucose), and 24-hour average glucose levels as determined by CGM.
Statistical Methods
For each of the variables of interest, there were up to 20 observations for each patient. Summary measures for each patient were computed as variable means for continuous variables and percent occurrence for binary variables. Both types of summary measures were treated as continuous. CSII programmers were compared to nonprogrammers on the summary measures using 2-sample t tests with adjustment for heterogeneity of variance. Pairwise Pearson correlations were computed. Statistical significance was taken as a 2-sided P value <.05. Calculations were performed in SAS ver. 9.3.
Source...