Analysis of “Comparison an Electronic Glycemic Management System Versus Provider Managed Subcutaneous Basal Bolus Insulin Therapy in the Hospital Setting”

Both hypo- and hyperglycemic excursions in hospitalized patients with diabetes are related to adverse outcome, prolonged hospital stay and increased cost. Therefore, current guidelines aim to safely achieve glycemic targets by recommending subcutaneous basal-bolus and correction insulin treatment for non–critically ill inpatients. For patients on the general ward, glucose targets between 140-180 mg/dl are recommended in most cases. ¹ Due to its complexity regarding ordering doses, correction schemes and the logistic challenge of blood glucose measurement and administration of insulin within appropriate time, basal bolus insulin treatment is still a challenging task and requires excellent interdisciplinary cooperation of all health care professionals. ² For this reason, computerized decision support systems have gained importance in recent years and their clinical performance has been investigated in several studies in different settings such as intensive care medicine, emergency department and general ward.^3-7

Aloi et al evaluated the safety and efficacy of basal bolus subcutaneous insulin therapy managed by providers, compared to a nurse-directed electronic glycemic management system (eGMS) using Glucommander in a retrospective observational crossover study in 993 non critically-ill patients across 9 different hospitals. ⁴ Mean blood glucose, number of hypoglycemic events less than 40 mg/dl and 70 mg/dl and the percentage of blood glucose in target (140-180 mg/dl) before (BGM), during (DGM), and after (AGM) the use of eGMS were compared. The authors concluded that using eGMS results in better glycemic control with less hypoglycemic events compared to provider managed basal-bolus insulin therapy (before and after eGMS treatment). Although the authors stress some of the limitations of the study like the retrospective, observational character, there might be further concerns that some of the findings need to be questioned.

The authors stated correctly that using a crossover design allowed a direct comparison between provider and eGMS managed therapy for the same patients. However, this design with lack of defined duration per treatment period can add to a treatment bias. It is not described who decided when to switch from provider-managed therapy to eGMS therapy and back. Usually patients are more ill at admission accompanied by hyperglycemia, higher degree of insulin resistance and greater insulin requirements. Therefore, the varying duration of treatment periods (BGM 1.2 vs DGM 4.5 vs AGM 4.2 days) might strongly influence the results. First, the duration of BGM is very short which might potentially simply not allow to establish glycemic control in such a limited period of time during which it can be assumed that patients are also in a more critical state than later on during the course of hospitalization. Second, it is important that DGM and AGM phases are rather comparable in duration. One might speculate that DGM was stopped to allow for initiation of another insulin regimen as discharge therapy, but obviously basal-bolus insulin was maintained, which would rule out this assumption. So the open question why eGMS was stopped remains.

The authors describe that during eGMS treatment initial insulin dose was either ordered by the nurse based on body weight (0.3, 0.5, or 0.7 U/kg total daily dose) or a custom initial dose could be set by providers. For BGM and AGM phases, insulin dose was directed by providers utilizing a computerized basal/bolus order set, using either a body-weight-based approach or an individualized dose based on provider’s judgment. However, the number and type of approaches chosen and the actual initial dose used has never been mentioned. Also, there is neither indication of the total daily dose nor of the distribution into basal and bolus component during the different treatment phases. A comparison of glycemic control without relating these data to starting insulin dose and insulin dose development over time seems difficult. To obtain a clear picture of user acceptance of the insulin dose suggestions, it would have been important to include the percentage of accepted dose suggestions in the DGM phase and to describe how acceptance evolved over time. Currently, eGMS dose suggestions seem to be followed in 100% of the cases, which seems to be rather unrealistic.

In addition, the authors do not describe what happened to potential preexisting antihyperglycemic medication (oral/injectable non-insulin agents) during the inpatient phase. Especially in patients on combination therapy (insulin plus other agents) an initial worsening of glycemic control would be assumed if hypoglycemic agents would be stopped upon admission and insulin dose would not be adjusted accordingly. This effect would then add to worsen glycemic control during the BGM phase and show improvement in the DGM phase when insulin finally was sufficiently (up-) titrated. To avoid some of the issues mentioned above, a randomized controlled trial (RCT) design probably may have been a less biased method to assess the influence of eGMS on glycemic control. Especially when considering that the initial titration phase is most crucial, an RCT design would have been favorable to see the efficacy of eGMS during the initialization phase.

It is also unclear how many patients had a diagnosis of preexisting diabetes and how many of these were already treated with insulin. It is well known that especially patients with long-standing type 2 diabetes are at the same risk of hypoglycemia as patients with type 1 diabetes. A more detailed description of the population (diabetes type, diabetes duration, preexisting insulin therapy and dose, body mass index, kidney function, preexisting antihyperglycemic medication) and possibly additional stratification for relevant factors (preexisting diabetes, diabetes duration, preexisting insulin therapy, HbA1c, kidney function, admission diagnosis, ward) would be of interest to make the results more generalizable. The results of the assumed eGMS effect on glycemic control and hypoglycemia risk should also be discussed especially with regard to similar studies in the field using either paper-based algorithms^8-11 or electronic decision support systems. ⁷

Overall, the authors made a strong case that eGMS haspositive impact on glycemic control in inpatients with diabetes requiring subcutaneous insulin therapy. Due to the limitations of the study it is however not possible to broadly recommend the use of Glucommander based on these results. The findings should be confirmed by a randomized controlled trial.