Abstract
Many current insulin pumps include bolus calculators to allow easy calculations of insulin doses. In addition, modes of bolus delivery can be chosen such as square-wave and dual-wave boluses, which are frequently used for high fat meals. Carbohydrate counting plus correction allows for added meal and snack flexibility and perhaps leads to better glycemic control compared with fixed meal content or insulin sliding scales, 3 but carbohydrate counting does not account for meals with higher amounts of fat and/or protein. Pizza is one common high fat meal in today's diet. Significantly, data indicate that relatively few (6.5%) youth with diabetes meet the American Diabetes Association goal for <10% of energy as saturated fat 4 or reach American Diabetes Association target A1c levels (44% of 3,947 youth with T1D). 5
There is emerging evidence indicating that fat and protein contribute to glycemia, which includes delayed gastric emptying, increased free fatty acids and insulin resistance, dysregulated postprandial glucagon release, and the decreased insulinotropic effects of glucose-dependent insulinotropic polypeptide in diabetes patients. The most important effect of fats on glycemia is due to delayed gastric emptying, which causes an increase in peak time and amplitude. 6 Generally, this effect reduces immediate postprandial glycemia but may result in late hyperglycemia. In a recent study, a fat (20 g)/water preload decreased postprandial glycemia, with no significant reduction in peak rise in glucose, but did cause higher glucose concentrations at later time points in type 2 diabetes. 7 Conversely, in T1D adolescents consuming a higher fat meal (36 g), the postprandial glycemic response was significantly flattened by slowing gastric emptying. 8 When protein is consumed together with dietary fat, the combination of delayed gastric emptying plus the insulinotropic effect of some amino acids can cause prolonged hyperinsulinemia (in healthy volunteers without diabetes), which ultimately has been shown to reduce insulin sensitivity via decreased glucose uptake in muscles and increased glucose uptake in adipose tissue through a change in glucose transporter type 4 9 and down-regulation of insulin receptors. 10 However, this mechanism has not been shown in T1D and likely contributes less to hyperglycemia in these patients than in healthy adolescents.
Recent studies in youth with T1D indicate that many adolescents do not count carbohydrates accurately. 11 How accurate carbohydrate estimation needs to be is seen with recent data suggesting that estimates within 10–15 g may be sufficient. 12 Moreover, it may be that precision (or systematic consistency in counting carbohydrates) may be more important than accuracy because if people are precise then a carbohydrate ratio adjusted to this inaccurate but precise carbohydrate counting will result in bolus insulin matched to carbohydrates. 13
Although carbohydrate counting and the effects of fat/protein on glycemia are frequently taught and observed as part of T1D care, few data exist on how to calculate insulin boluses for mixed meals in diabetes. Therefore, Pańkowska et al. 14 provide provocative data that question whether a simple bolus for the carbohydrate content of a pizza meal is sufficient to maintain near euglycemia compared with an algorithm in which a normal-wave bolus is given for carbohydrates and also a square-wave bolus for fat/protein content. For patients and diabetologists engaged in improving upon current systems for determining bolus insulin dosing, these data have potential clinical application.
In response to the complex but common clinical problem of how to dose insulin for a mixed meal, Pańkowska et al. 14 designed an inpatient fixed meal study in 23 adolescents with T1D and hypothesized that a meal bolus algorithm accounting for both carbohydrates and fat/protein will maintain euglycemia better than a bolus for only carbohydrates. Mixed meals of carbohydrate, fat, and protein—in varying amounts—are what patients actually eat, and this study attempts to address this complexity with a two-part algorithm.
Dosing for carbohydrates was based on carbohydrate units (10 g=1 carbohydrate unit), and the second arm of the study received an additional unit of insulin for each fat-protein unit (FPU), defined as 100 kcal of fat and protein. Subjects' blood glucose levels were measured over a 6-h period, and as all subjects consumed a fixed meal, subjects received either 4.5 units as a normal-wave bolus to cover carbohydrates or these 4.5 units to cover the carbohydrates plus an additional 4 units as a 6-h extended wave bolus (8.5 units total compared with 4.5 units). Not surprisingly, the arm that received more insulin had a lower mean glucose at 2, 4, and 6 h (and also more hypoglycemia [33% compared with 0%]).
Such a study is difficult to design and execute, and the authors are to be commended for their efforts, although some questions exist on the study design that could affect interpretation of the data. As better matching of insulin dose to consumed food is an underlying goal of the investigation, it is uncertain why all subjects used the same carbohydrate and FPU ratios, especially given baseline differences within study subjects. Of some concern are between-group differences in body mass index (1.15 SD score), low-density lipoprotein cholesterol (19 mg/dL), and high-density lipoprotein cholesterol (35 mg/dL), suggestive of between-group differences in insulin sensitivity. Also, although the between-group difference in change in glucose from baseline is emphasized, the control arm's mean baseline glucose was 40 mg/dL higher than that of the intervention arm, although not quite significant (P=0.054). Moreover, different dosing regimens—normal-wave bolus for the control group versus dual-wave bolus for the intervention group—were used even though a dual-wave bolus by itself has been shown to improve glycemic control following a pizza meal. 15,16 In these studies, the same amount of insulin was given for a high fat meal with the difference being the method of bolus delivery (normal- vs. dual-wave); no increases in hypoglycemia were reported in these studies. 15,16 So, it is not entirely clear if the results of the study of Pańkowska et al. 14 are due to the new dosing algorithm and/or the method of insulin delivery and increased amount of bolus insulin. However, the studies must start somewhere, and at some point such algorithms must be “user-friendly” to translate to clinical practice.
In their discussion, the authors identify potential mechanisms by which fat/protein lead to postprandial hyperglycemia: delayed gastric emptying, increased free fatty acids and insulin resistance, conversion to amino acids and subsequently gluconeogenesis, and elevations in glucagon (not supported by their data), all of which are testable in future studies. The authors acknowledge limitations in their study but assert that their data shed new light on the strategy of functional insulin therapy and encourage a critical dialogue on inclusion of fat/protein in bolus decisions. Intriguingly, they state that with current pump technology an FPU could be programmed with a subsequent square-wave bolus much as normal-wave boluses are currently delivered based on carbohydrates. Prior to such adoption, additional studies would be required in both inpatients and outpatients to determine such dosing's safety over a 6-h period (or longer), especially in people with T1D who exercise; a major limitation is the 33% frequency of hypoglycemia over 6 h. Individualized adjustment of these ratios should improve the effect on glycemic outcomes.
In addition to the daunting challenge of calculating FPUs when it is questionable whether carbohydrates are counted correctly, 11 many patients might be hesitant to use a bolusing method with a one in three chance of hypoglycemia within 6 h. However, these challenges should not discourage future research to refine bolus insulin dosing algorithms, especially given advances in technology and information that could include use of smart pumps and indexes of food macronutrient content. Although such efforts are potentially translatable to current patients with T1D, future applications include versions of a hybrid-closed loop system in which premeal bolus dosing has been shown to improve glycemic excursions. 17 Further precision for these dosing algorithms is required prior to widespread adoption as it could be argued that the current study merely demonstrates that if additional insulin is dosed, then of course lower blood glucose levels (and more hypoglycemia) result.
Where the balance lies for each patient and diabetes provider for additional complexity versus simplicity for bolus dosing decisions will likely remain an individual matter. A recent editorial 18 and letter 19 in this journal debated the importance of potential misdosing of 1–2 units of insulin with airline travel and its effects on glycemic control and safety; in comparison, this proposed method adds 4 units over a 6-h period. Although achieving euglycemia 24 h/day is not attainable with current technology, an algorithm to better individualize bolus insulin with meals, if further studies both inpatient and outpatient support this, would advance current care. In sum, Pańkowska et al. 14 provide intriguing data suggesting that better dosing decisions for mixed meals are possible, although further research is required to refine such algorithms and demonstrate effectiveness in outpatient clinical settings.