Response to the Letters to the Editor

The two Letters to the Editor comment on an article that appeared in the December 2009 issue of Hospital Pharmacy (Dickerson RN. Glycemic control during critical illness: tight or not? Hosp Pharm. 2009;44(12):1142–1148). It is good to have dialogue regarding the best way to manage hyperglycemia in the intensive care unit (ICU), as this is an important issue that many pharmacists are intimately engaged in at their respective institutions. Walker-Renard described her difficulties with a paper-based protocol, which ultimately led to the use of a commercially available, computer-based, glucose management system. We had just the opposite experience with a different computer-based glucose management tool in our medical ICU a few years ago: several patients developed severe hypoglycemia and some of the insulin dosage recommendations were disregarded due to clinical suspicion that it would be too dangerous (eg, increase the insulin infusion rate by 16 units/h). As a result, our institution abandoned its use. It is encouraging that this technology and associated computer algorithms are improving with good results for some institutions. I would encourage the author to continue to accrue data and publish the results in a rigorous, peer-reviewed journal.

I am concerned by the comments of Szumita and colleagues that I advocate a “nonvalidated ‘fixed-dose’ insulin infusion.” I agree that a fixed-dose algorithm is an ineffective method for managing hyperglycemia, but our algorithm is not a fixed-dose insulin infusion as suggested by the authors! With close inspection of Figure 1 in the Hospital Pharmacy review ¹ (our regular human insulin [RHI] infusion order form used in clinical practice), it is evident that our algorithm is dynamic. For example, if a patient is started on an RHI infusion at 2 units/h and his blood glucose concentration (BG) 1 hour later is 210 mg/dL, the algorithm calls for the infusion to be increased by 3 units/h (to a final rate of 5 units/h). If the patient's next hourly BG is 165 mg/dL, the infusion rate is to be increased by 1 unit/h (to a final RHI infusion rate of 6 units/h). If the next BG is within the target range, no change in the RHI infusion is warranted. If the BG is too low, a reduction in infusion rate is indicated per the protocol/algorithm. Using this methodology, the RHI infusion is tailored to the individual's glycemic response (and is not a fixed RHI dosage).

Another erroneous conclusion from the Boston pharmacist group was that our algorithm was “nonvalidated” and lacked “various metric assessment variables.” I am perplexed by this erroneous statement as we have evaluated this algorithm in our original paper ² from 2008 using “metric assessment variables.” The original publication was referred to and cited at least a couple of times in the Hospital Pharmacy article. Szumita and associates should have read the full details from the original publication and not based their criticisms on a mini-review of the topic. Additionally, we have recently learned that this 2008 paper has been selected for the John M. Kinney International Award for the most outstanding original work relevant to the field of General Nutrition published during 2008-2009 in Nutrition: The International Journal of Applied and Basic Nutritional Sciences, with the award ceremony to be conducted at the February 2010 meeting of the American Society for Parenteral and Enteral Nutrition meeting in Las Vegas. Our opinion, with about 4 years of experience using this algorithm, is that it works very well for our trauma, surgical, and thermally injured patient population receiving continuous enteral or parenteral nutrition as long as the patient does not have acute kidney injury or chronic renal insufficiency. ³

The final concern raised by Szumita and associates is that we did not report our compliance rate, yet criticized their insulin infusion protocol violation rate of 32%. ⁴ Unfortunately, we did not analyze those data in our original study, as we naïvely did not think it was a significant problem with our algorithm. ² We are contemplating whether we, as well as others, would find it useful to ascertain and publish that information by reexamining our raw data as we recorded hourly BG and hourly RHI infusion rates for each patient. Despite our lack of published data regarding errors by nursing personnel in our study, I would still contend that a 32% error rate in the determination of prescribed insulin infusion rate by nursing personnel as reported by Cyrus and coworkers ⁴ is a significant problem likely related to the complexity of their algorithm despite overall good glycemic control.

As for the need for complexity for a safe and effective insulin infusion protocol, I would recommend that the reader examine our 2008 paper in addition to other published protocols, assess their extent and duration of glycemic control, incidence and severity of hypoglycemia, presence or absence of specialized nutrition support, and the patient population and then draw their own conclusions. The “best method” as well as the “optimal” blood glucose target range for glycemic control during critical illness will likely continue to evolve throughout the upcoming decade. It is hoped that the reader is as passionate as those involved in this discussion regarding glycemic control during critical illness. It is only through fruitful informed discussions and further clinical research that patient care will continue to advance.