Analysis of insulin pump settings in children and adolescents with type 1 diabetes mellitus

Methods This retrospective study included patients aged 1 year who were using the Medtronic pump device. Data from the insulin pumps including number of blood glucose (BG) tests per day, basal and bolus insulin parameters, carbohydrate ratio (CR) and insulin sensitivity factors (ISF) were averaged over 14 days for statistical analyses. Anthropometric data and recent HbA1c was also recorded.

Insulin pump therapy (continuous subcutaneous insulin infusion, CSII) has become increasingly established as a routine therapy for the management of type 1 diabetes mellitus (T1DM) in children and adolescents (1). While studies have shown that CSII provides potential advantages over injection regimens such as greater lifestyle flexibility, reduced risk of hypoglycemia, and improved glycemic control (2)(3)(4)(5), the target glycosylated hemoglobin A1c (HbA1c) of <7.5%, as recommended by the International Society for Pediatric and Adolescent Diabetes, is still not achieved in many young patients (6,7).
Effective insulin pump therapy requires regular adjustment for optimization of settings for basal rate delivery and bolusing for carbohydrate intake and hyperglycemia correction which are dependent on age, eating and exercise patterns, and other lifestyle factors (8). Consensus guidelines in pediatric cohort give extensive and useful information on management of diabetes in those on insulin injections (9). Although consensus guidelines have been put forth for standardization of initial pump settings in adults, there appear to be no such guidelines in the pediatric and adolescent population (10). Current strategies for tailoring insulin pump settings in children tend to rely on empirical clinical experience rather than on clinical evidence (1).
Evidence suggests better glycemic control in insulin pump users to correlate with several factors including higher number of daily boluses (2,11,12) and frequency of mealtime boluses (13)(14)(15). One study suggested a higher number of basal rates to be associated with better HbA1c (16) and another with increased duration of temporary basal rate usage (17). Adherence with daily tasks of diabetes self-management, including frequency of blood glucose (BG) measurements, has also been associated with better glycemic control (2,17,18). Furthermore, various tools have been designed to accurately assess adherence and high usage of the existing technology (19). As clinicians, we are also aware of the huge role that psychosocial support systems play in the compliance and overall outcome especially among adolescents with type 1 diabetes (20). However, inconsistent associations have been found between other insulin dosing variables and HbA1c (11,17,18).
Thus, there is a critical need for more data to guide optimization of bolus settings for carbohydrate intake and correction doses across age groups with the aim of helping more children and adolescents to meet their glycemic targets. Some approaches have been presented in the literature (12,21,22) but anecdotally a wide variety of approaches are practiced across pediatric units. Therefore, to attempt to fill that gap in knowledge, the aim of this study was to analyze the pump settings that children and adolescents receive in a large tertiary pediatric diabetes center and explore the relationship between insulin dosing patterns (basal and bolus settings), age, and glycemic control.
Specifically we aimed to: i. Describe the basal and bolus insulin dosing patterns in children and adolescents on CSII; ii. Compare real-life insulin pump settings such as carbohydrate ratios (CRs) and insulin sensitivity factors (ISFs) to initial estimates based on the '500' rule and '100' rules, respectively; and iii. Investigate the relationship between insulin pump settings and HbA1c.

Study design
This retrospective study was conducted at the Diabetes Clinic of The Children's Hospital at Westmead in August 2013. The study involved collecting and analyzing insulin pump data from patients who used Medtronic insulin pumps (Medtronic Inc, Northridge, CA, USA) (97% of the clinic pump users) and visited the clinic in the 12 months from August 2012 to August 2013. Pump downloads from the most recent clinic visit for each patient were extracted from the Medtronic CareLink software and averaged over 2 weeks for statistical analyses. There were a total of 641 patients who attended the diabetes clinic in this period of 12 months, of whom 300 were treated with insulin pump therapy.

Subjects
Pump download data were analyzed from the most recent clinic review within the last 12 months (14d download) for all children and adolescents aged <18 yr who were using the Medtronic pump device. Patients who had experienced T1DM for <1 yr were excluded from the study to avoid confounding data from a remission phase. There were 292 patients who met these criteria and all had 14-d pump download available which was used for the analysis. All subjects used a rapid-acting insulin analog in their pump, either insulin lispro or insulin aspart.

Study variables
Clinical data retrieved included age, sex, height, weight, HbA1c, T1DM duration, insulin pump duration, BG level, number of BG tests per day, total daily insulin, basal and bolus insulin parameters, CR, and ISF. Standard deviation scores (SDS) for height, weight, and body mass index (BMI) were calculated using the US CDC 2000 growth standards.
Standardized criteria for pump settings are used in our unit at pump initiation. Total daily dose (TDD) is reduced by 20% from injection doses or to 1 unit/kg/d, whichever is the lowest. Basal rates are calculated from 45 to 50% of the calculated TDD divided into a biphasic pattern according to age with morning basal rates rising from 01:00 hours in adolescents or 06:00 hours in preadolescents. Insulin to CR and ISF at pump initiation are calculated by the 500 rule (500 divided by TDD = number of grams of carbohydrate covered by 1 unit of insulin) and 100 rule (100 divided by TDD = number of mmol/L that 1 unit of insulin will lower the BG), respectively (the 100 rule is referred to as the 1800 rule when BG level is measured in mg/dL rather than mmol/L). Initial adjustments to pump settings are performed weekly by the diabetes education team based on pump and glucose meter downloads until stable and all patients received comprehensive clinical review for 3 months. Adjustments in the bolus variables are usually made in increments of 10-20% if needed. The maximum percentages of variations in CR and ISF were calculated using the formula: [(highest value − lowest value)/lowest value] × 100%.

Statistical analysis
Statistical analyses were performed using the IBM Statistical Package for the Social Sciences (spss) software 21. For descriptive statistics, patients' data were divided into three groups according to their ages: preschool (<6 yr), preadolescent (6 to <12 yr), and adolescents (12 to <18 yr) on the basis that these age groups have previously been shown to have physiological differences in patterns of insulin requirements. (11,23,24). The normality of the distribution of each study variable was assessed using the Kolmogorov-Smirnov statistics. Results for descriptive statistics were presented as mean ± standard deviation (SD) for normally distributed data and median (interquartile range, IQR) for non-normally distributed data. One-way anova and Kruskal-Wallis tests were performed as appropriate for normally distributed and nonnormally distributed data, respectively, to identify the differences in insulin dosing patterns across three age groups. Significant results were followed by pairwise comparisons with Bonferroni correction to examine differences between age groups. p-Values of <0.05 were considered statistically significant. Multiple regression models were used to examine variables associated with metabolic control. Independent variables that were considered for the multiple regression models included age, diabetes duration, pump duration, basal and bolus insulin variables, TDD, number of BG tests per day, CR, and ISF parameters. An independent variable within each highly correlated independent variable pair was excluded from the regression models to ensure that the assumption of non-multicollinearity was met. Also, the normal probability plots of the regression standardized residuals were examined as part of the multiple regression analysis to examine for violations of homoscedasticity.

Ethics
This study received research ethics approval from the Sydney Children's Hospitals Network Human Research Ethics Committee on 5 July 2013.

Results
Of the 641 T1DM patients who visited the clinic from August 2012 to August 2013, 300 were pump users. Totally, 292 patients (144 males and 148 females) were on Medtronic pump and had diabetes for >1 yr. Pump download data (14-d download) from this group were analyzed for the purpose of this study.

Bolus data
The median (IQR) number of daily boluses administered by the entire study population was 6.3  (4.9-8.0). More boluses per day were administered in preschool subjects 8.0 (6.5-9.3) and preadolescent subjects 7.0 (6.0-9.0) compared with adolescents 6.0 (4.0-7.0) (p < 0.01). There was a significant improvement in HbA1c with an increase in the number of boluses, with those receiving <4 boluses per day having a median HbA1c of 9.1 and those having >12 boluses per day of 7.8 (p < 0.001). An increased number of boluses was also associated with younger age and shorter diabetes duration (p < 0.001).

Bolus wizard variables
Mean CR and ISF ratio absolute values and per kilogram decreased with age as expected ( Table 2). Individual meal time CRs within each age group did not differ significantly, although there was a trend to stronger CR at breakfast in the two younger age groups (actual breakfast CR was 70% of the calculated CR based on the 500 rule in preschool age group and 73.5% in the preadolescent age group). For all age groups, the average CR of the patients was lower than the calculated 500 rule while the average ISF was higher than the calculated 100 rule ( Table 2). Our average patient settings equated to a '430 rule' for CR settings and a '120 rule' for ISF.

Basal data
Total daily insulin dose, basal insulin dose, and percentage of basal insulin per TDD were found to increase with age ( Table 2). Basal insulin patterns showed diurnal variation and differences with age groups (Fig. 1). The proportion of basal insulin received from 00:00 to 06:00 hours was significantly higher in adolescents than preadolescent children but was significantly lower from 18:00 to 00:00 hours compared with preschool and preadolescent children (p < 0.05).

Glycemic control
Despite a large number of variables considered for inclusion in the multiple regression analyses, only a small number was found to be significantly associated with glycemic control. Multiple regression analyses showed that lower HbA1c values were associated with higher number of daily boluses, greater number of BG tests per day, lower average CR/500 rule ratio (indicating stronger carbohydrate bolusing), and higher average ISF/100 rule ratio (indicating weaker corrections), adjusted for age (R 2 = 0.22; p < 0.01) ( Table 3).

Discussion
Insulin pump therapy is being increasingly used for management of children and adolescents with type 1 diabetes. Current insulin pumps lack automated dose adjustment based on feedback control, so pump settings need on-going adjustment by pump users in conjunction with their diabetes health professionals. The opportunity of having electronic data from modern insulin pumps enables the patients and the diabetes management team to objectively evaluate and review administered insulin treatment (25) and to use it in statistical analyses for study purposes (1). The main focus of this study was to assist in filling the void in the current evidence on bolus wizard and basal settings in pediatric population as well as to look at their associations with glycemic control.
This study showed that patterns of bolus and basal insulin dosing varied between preschool, preadolescent, and adolescent patients. Our findings on the  patterns of bolus insulin use closely coincided with results from previous studies where adolescents were shown to administer significantly fewer number of boluses per day compared with patients from younger age groups (11,12,17). Literature suggests that this pattern is related to decreased parental involvement and to missing boluses for weight control or to avoid episodes of hypoglycemia (12,17), but may also partly reflect age-related eating patterns. Higher number of boluses was significantly associated with lower HbA1c in our study as noted by earlier studies (11,17). Adolescent data reveal a higher percentage basal rate compared with preschool and preadolescent patients, probably reflecting automatic basal delivery with reduced total bolus delivery and/or increases in basal rates made to compensate for inadequate bolusing. In patients with good adherence, this could also represent a state of relative insulin resistance resulting in the need for increased basal insulin delivery. However, a higher percentage of basal insulin was associated with a small increase in BMI-SDS. Frequency of mealtime insulin bolus has been studied as a measure of adherence as well as a predictor of future HbA1c (13)(14)(15). In contrast to the findings of previous studies wherein poorer glycemic control was observed in adolescents who took fewer daily boluses and BG tests per day (12), HbA1c in adolescents of this study was not significantly higher than those from the younger age groups despite adolescents having fewer daily boluses and BG checks. This difference may be due to stringent correction applied to multiple comparisons in this study. It may also signify the presence of other unassessed lifestyle factors such as exercise which may have contributed to glycemic control. Median HbA1c among all children and adolescents with type 1 diabetes attending the clinic for the same time period was 8.2% (IQR: 7.5-9.2%) which was not different to the patients enrolled into our study, suggesting no selection bias.
In order to deliver tailored doses of insulin, modern insulin pumps allow users to program multiple CR and ISFs for different periods of the day for calculation of correction boluses (8). There are very few data in the literature exploring the appropriateness of the rules of thumb used and comparing to real-world values. At the initiation of CSII therapy, the diabetes team at our clinic uses the 500 rule and 100 rule to estimate CR and ISF, respectively, as a starting point from which adjustments are made over time to suit the individual. Based on our findings, the 500 rule overestimates the real-life CR values (i.e., a stronger CR is used in reallife), whereas the 100 rule underestimates the average ISF used in real life (i.e., a weaker ISF is used in reallife). Our average patient settings equated to a '430 rule' for CR and a '120 rule' for ISF. A previous study carried out among 154 children with type 1 diabetes found similar results in young children, <7 and 7-13 yr (16). In this study cohort, the subjects were designated into eight groups based on their sex, age, and pubertal status. The youngest (<7 yr) subgroup of children with T1DM had a lower mean CR [termed CR-prediction factor (PF) in this study] of 316 and higher mean ISF-PF of 144 (or 2588 if BG level in mg/dL) compared with all other groups and the theoretical adult values of 450 for CR and 100 for ISF (or 1800 if BG level measured in mg/dL) (21). These authors also reported a lower CR at breakfast in the younger age group compared with other groups. We also looked at the variability in meal time CRs within each group and found no statistical difference, although there was a trend to a stronger carbohydrate bolusing at breakfast time.
Total daily insulin dose and basal insulin dose increased with age as expected. As with previous studies that investigated the circadian profiles of basal insulin between different age groups (23,24,26), we showed an increase in basal insulin requirement in adolescents during the early morning hours, corresponding to the dawn phenomenon as well as an increase in basal insulin requirement in preschool children during the late evening, corresponding to the dusk (or reverse dawn) phenomenon. Reasons for these phenomena remain unclear, although some researchers have proposed that the differences in growth hormone and cortisol secretion patterns between children and adolescents may be the cause for predawn insulin resistance (27). There is also the possibility that the dusk phenomenon relates to precautionary conservatism in meal bolusing in the evening. These findings may also be biased by an innate tendency to adjust the rates by the managing team in keeping with the findings of the earlier studies or unit practice, as mentioned above.
In our study, better glycemic control was associated with a higher number of daily boluses, and greater number of BG tests per day which has also been reported in other studies (2,17). Lower HbA1c was also associated with lower average CR/500 rule ratio and higher average ISF/100 rule ratio adjusted for age (R 2 = 0.22; p < 0.01), which has not been evaluated in any other pediatric cohort. This knowledge may be helpful in guiding future practice to achieve better glycemic outcomes. This study did not find any relationship between number of basal rates and glycemic control, although a previous study by Nabhan et al. demonstrated a higher number of basal rates to be associated with improved glycemic control (16). A possible reason for this finding is that there was a very small range of the data, presumably because of the established prescribing policy within the unit. There was also no correlation between total daily insulin dose/kg and HbA1c in our study, as also noted by O'Connell et al (18).
Limitations of this study include the retrospective design which did not allow for evaluation of other factors which may be important in the understanding of insulin dosing in children and adolescents, such as family involvement and patients' satisfaction toward CSII therapy. Also, data on bolus adherence, number of hypoglycemic episodes, or instances of diabetic ketoacidosis were not available as reliable data through the pump downloads. A few earlier studies have shown minimal increase in risk of ketoacidosis among the pump group as compared with MDI group and hence that would have been useful information to gather (4,5). This is a single-site study and hence the findings could have been biased by standardized practices used for adjusting the insulin pump settings by the management team. Nevertheless, we believe that the data represent a useful snapshot of the pump settings that have evolved in a large tertiary center and this will allow readers to compare it with their own practices, allow them to consider whether other approaches to pump settings may be superior, and may prompt further studies of these issues. It is also possible that by using linear regression models for analyzing we may have overlooked other factors that have non-linear associations such as patient motivation to seek medical advice in between appointments, family support measures, etc. The main strength of our study is the number of subjects enrolled and the corrections applied for multiple comparisons for analysis.

Conclusion
For CSII therapy in children and adolescents, individual adjustment and frequent re-evaluation of insulin dose settings are necessary for effective management of type 1 diabetes; however, it is useful to have a guide to common settings that are applicable to each age group. Our data suggest that stronger CR settings and weaker ISF settings are needed in the real-world compared with established rules of thumb and are associated with better glycemic control. Our data also confirm previously reported circadian profiles of basal insulin delivery among the preschool, preadolescent, and adolescent individuals. Future studies should be designed to determine whether stronger CR settings and weaker ISF settings at insulin pump initiation predict future improvements in HbA1c. These will further help in establishing guidelines for insulin pump settings at CSII initiation in the pediatric population.