Commentary on July 19 New York Times article on glucose monitor accuracy


2009-08-14 16:34:46

By: Richard Hellman, MD, FACP, FACE

As readers of this editorial series know, there is considerable concern from many quarters that the lower-than-optimal accuracy of glucose meters in current use, both in the inpatient and outpatient setting, is a cause of errors in insulin dosage, with resultant morbidity, and possibly mortality. In an article in Clinical Chemistry in 2008, Dr. David Sacks and his colleagues cited the evidence for the widely held, but not universal, concern that the accuracy of inpatient meters used in the United States needs improvement.


An article in the July 19, 2009 issue of the New York Times, by Gardiner Harris, highlighted this issue. In the article, the reporter discussed a recent letter from Dr. Margaret Hamburg, the Commissioner of the FDA, which stated that the FDA is considering requiring a higher level of accuracy and precision for glucose meters than is the standard today (that 90% of the readings be within plus or minus 20% of the reference value).


The FDA letter to AACE was in response to a letter from three key members of AACE, including the President and two Past Presidents. I am proud to say that I was one of those who signed the letter, which asked the FDA to consider both raising the standards of accuracy and precision that meter manufacturers would need to meet for approved products, and to increase the quality of the patient education information given by the manufacturers so patients could be more aware of the special features and limitations of their glucose meters.


The NY Times article cited the opinion of the FDA’s Dr. Alberto Gutierrez, the deputy director of the Center for Devices and Radiological Health Office of In Vitro Diagnostic Device Evaluation and Safety. Dr. Gutierrez’ opinion was that the current glucose meters in common use were not designed for inpatient use, and are not sufficiently accurate for the hospital treatment of critically ill patients.


In addition, the Times article cited some of the drug interferences that commonly occur, such as those due to Tylenol or vitamin C, and the problems with peritoneal dialysis when using Accu-Chek meters.


Interestingly, the NY Times article cited the data by government researchers which showed that when comparing tests from five different popular meters, the results varied by as much as 32%. They also cited a most unusual and sophisticated high school science project that studied seven different glucose monitors and found that two of the meters from the same company varied by an average of 62 mg%. Although the study was done by a high school student, the faculty advisor was a highly trained, well-respected academic clinical investigator, and the study appeared to be at a much higher level than one would expect from a high school project.


The NY Times article concludes with a quote from Dr. David Sacks, an Associate Professor of Pathology at the Harvard Medical School: “Insulin is a dangerous drug, and if someone makes the wrong decision about its use because of a bad test, they could die.”


It should be interesting to see the result that stems from the increased attention of the NY Times article to the topic of the accuracy of glucose meters. I hope the focus remains on improving decisions for the person with diabetes by improving the quality of data used for the decision. I think this will increase the value of glucose testing and go a long way to improve patient safety.



    1. Standards Might Rise on Monitors for Diabetics by Gardiner Harris, New York Times, Sunday, July 19, 2009, pages 14 and 20




Safety in the ICU - Is Bedside Glucose Testing Safe For Our Patients?


2009-02-25 09:56:39

By: Richard Hellman, MD, FACP, FACE

In my recent editorial on the NICE-SUGAR study, I raised the question as to whether the methods commonly used to measure glucose levels at the bedside may be the source of error that may pose a hazard to the patient. A very important article by Scott, Bruns, Boyd, and Sacks, just published on-line in Clinical Chemistry1, reviews this important point and adds important new information to the excellent review by Dungan in 2007 in Diabetes Care2.


For several years, the clinical chemists and pathologists have been exploring the limitations of bedside glucose testing. Alterations in hematocrit and oxygen concentration may affect the result. Variations may be introduced by the site used to obtain the sample and whether the area is a site of low blood flow. In some cases, serious errors in insulin dosage have occurred when the bedside glucose method was not glucose specific and non-glucose sugars, such as icodextrin, used in peritoneal dialysis have caused the bedside readings to be falsely elevated, leading the clinicians into believing that more insulin was warranted, resulting in insulin overdosing and patient injury and deaths.


But Scott and his colleagues reviewed the evidence that suggests that the recent tight glucose control meta-analysis that failed to find the beneficial outcomes reported by Van den Berghe may have resulted from failure to ensure that the method of measuring glucose levels at the bedside were sufficiently accurate so as to prevent incorrect decisions, leading to iatrogenic hypoglycemia. Their analysis may be completely correct. They also comment on an earlier study by Boyd and Bruns who carried out simulation modeling studies to simulate the effects of the lack of precision of the glucose meters on the chosen dose of insulin for two insulin administration algorithms.


Their findings were sobering. A simulated total analytical error of only 5% led to an incorrect insulin dose in 8-23% of cases, but a total analytical error of 10% lead to incorrect insulin dosing in 16-45% of cases. To put this into context, the FDA allows 20% error for package insert claims of glucose meters. In other words, the acceptable standard for accuracy and precision of these medical devices is well below what is safe for our patients in critical care situations.


On the other hand, there is evidence that with proper training, the most accurate and precise glucose meters can be used safely in many settings in hospitals, and yield highly accurate results in a most timely fashion, and be useful in settings such as post-open heart surgery etc. But it appears that, until the FDA provides a higher standard of accuracy and precision and begins testing each type of instrument to verify the accuracy, it will be left to the providers of care to do a careful analytic review of the methods used in their institution, and to properly train their personnel and verify that the results they yield are sufficiently accurate to be used in the care of the vulnerable patient.


In summary, it appears that there is yet another part of the system of care that needs to be not taken for granted, if we are to achieve normoglycemia in our patients more often, and to avoid unnecessary hypoglycemia. We need to recheck whether our glucose testing is up to the task at hand. Scott et al have written an excellent reminder of the importance of looking carefully at the tools we use for glucose monitoring.




1. Scott MG, et al. Tight Glucose Control in the Intensive Care Unit: Are Glucose Meters up to the Task? Clinical Chemistry 55:1 18-20 (2009)


2. Dungan K, et al. Glucose Measurement: Confounding Issues in Setting Targets for Inpatient Management. Diabetes Care 30:403-409, 2007





How Important is Overconfidence in the Generation of Diagnostic Errors?


2009-02-26 16:30:29

By: Dr. Richard Hellman

Few patients tolerate uncertainty and hesitancy in their physician. The confident physician generates respect, admiration, and a following from their patients, staff and colleagues. But when does confidence in their own diagnostic acumen become a liability, and overconfidence lead to premature closure of the diagnostic process, and diagnostic error?


The answer may surprise you. Unlike the physicians in the two “perceptive” specialties, radiology and pathology, who use primarily a visual based process of diagnosis and generally, but not always, have a lower rate of diagnostic errors, there are data which suggest that the diagnostic error frequency among most physicians in the non-perceptive specialties in outpatient settings is ~15%. This estimate may be conservative. For example, other data, from the Harvard Medical Practice study, showed diagnostic errors accounted for 17% of adverse events. Higher rates are also reported in autopsy-based studies.


In data published in the American Journal of Medicine’s Symposium on Overconfidence and Diagnostic Error in May 2008, Graber reports that the physicians interviewed assessed their rate of diagnostic errors as ~1%, much less than the measured rates of 15%. In an article by Berner and Graber, they report that in several studies, physicians readily acknowledged the high frequency of diagnostic error, but held to the rarity of their errors1. In short, the physicians measured believed that diagnostic errors were common – among other doctors. This is not surprising. 94% of surgeons believe they are in the top 50% of their profession.


The discernment with skill and confidence levels is apparent in the studies of Potchen2, who studied the relationship in board certified radiologists between physician confidence in their diagnosis and the correct diagnosis. They looked at those physicians who were in the top 20% of diagnostic accuracy, and those in the lower 20% group. The group with the highest level of confidence in their diagnosis was the bottom 20%.


The discordance between performance and overconfidence was particularly striking in studies with medical residents, who, in comparison to practicing physicians and medical students, had the highest disparity between their confidence (very high) and their performance (much lower than the practicing physician). The medical students in contrast, both performed poorly and were not at all confident in their diagnosis.


So why should we discuss the subject in this column at all? For starters, there is evidence from malpractice records that the most common type of malpractice is a diagnostic error that results in a permanent injury to a patient. There is also evidence, that we published in 2001 (Hellman), that in diabetic care, the most frequent association of a catastrophic medical error that led to death within 24 hours was a diagnostic error in a setting where there was a defective culture of safety, and inadequate systems were available to protect the patient against the consequences of the physician’s diagnostic error.


It is always easier to blame others for errors, and many people initially have an uncharitable view of why diagnostic overconfidence should exist among physicians. But a more reasoned approach can be found in an article in the AMJ symposium by Gordon Schiff3. In his article, he discusses the severe pressures on physicians and the inherent hazards of a system of care that makes it very difficult to get timely feedback regarding the effect or result of their diagnostic and/or therapeutic efforts on their patients. He points out that the lack of good correlation of care and inadequate feedback are two important factors that increase the risk that a physician will prematurely close the door on the correct diagnostic possibility.


If we are to reduce the frequency of diagnostic errors, it will take a combined effort, not just by physicians, but also payors, policy makers and the public to become aware of the need to be protective of processes and resources needed for physicians to maximize their diagnostic accuracy. It is equally important to avoid creating workplaces where the speed with which a physician practices is so rapid as to preclude a careful, analytical process of evaluating the patients’ diagnoses and then therapeutic needs. Although an experienced and skilled physician can often use pattern recognition to accurately diagnose the problems at hand, sometimes only a careful analysis approach to diagnosis will yield the correct answer.


While most of us realize that in an ambulatory care setting it is hard to spot a diagnostic “zebra” when we hear just hoof beats, there is less general awareness that even a common diagnosis may present in an uncommon way. When we are pressed for time, if we do not ask the extra question, to reflect further on “what else this could be”, even the best clinician can focus prematurely on the wrong patient diagnosis, and this can lead to injurious medical errors.


How then, can we reduce diagnostic errors? For starters, we need to develop work-flows that take into account how the stresses of a busy clinical setting can lead to cognitive errors. Decision support aids are useful, but are just a beginning. Proper follow-up often reveals facts that were not apparent at the initial visit and shows possibilities that were not completely explored initially.


On the other hand, it may be useful to have decision support software to automatically remind us that the new referral, whose demographics indicate that they were born in Japan, and who presents with progressive and more frequent weakness, may not only have a higher incidence of gastric cancer, but also may have a increased risk of hypokalemic paralysis associated with hyperthyroidism.


In addition, we need to construct systems of care in our own clinical settings, and to work more closely with the larger systems of care we interface with, so not only our staff are properly educated, but, wherever possible, we have contributed to the education of key staff in the hospitals or other facilities that we are in some of the time. The more we can extend our reach to help others get the level of expertise we want, the safer and more efficient the care of our patients will tend to be.


But, the most challenging task will be to deal with ourselves. We do have a right to be confident in our skills and diagnostic acumen. But that should not prevent us from regularly checking our own work and being more transparent in our clinical reasoning so our staff and colleagues will be educated by our logical analysis, and in the rare event of a slip or lapse on our part, more likely to spot the oversight before the error causes harm. A more realistic understanding of how we make diagnostic decisions, and when they are most sound, will allow us to be even more likely to achieve the level of diagnostic excellence we expect of ourselves.



    1. Burner E & Graber M. Overconfidence as a Cause of Diagnostic Error in Medicine, Am J Med 121(5A): S2-S23
    2. Potchen EJ. Measuring observer performance in chest radiology: some experiences, J Am Coll Radiol 3(6): 423-32
    3. Schiff G. Minimizing diagnostic error: the importance of follow-up and feedback, Am J Med 121(5 Suppl): S38-S42




When do Medicare Payment Policies Potentially Threaten Patient Safety? A Potential Example: Proposed Payment Policies for Measurement of 25-OH Vitamin D Levels


2009-02-26 16:32:48

By: Dr. Richard Hellman

A draft Local Coverage Determination (LCD) was issued by one of the regional CMS carriers this month on payment codes for calcifediol (25-OH Vitamin D-3) levels. In this LCD, they delineated the diagnosis codes that would be paid for if the 25-OH vitamin D level was assayed. The covered diagnoses included only chronic kidney disease, hypercalcemia, and disorders of phosphorous metabolism, osteomalacia, and active rickets.


This highly restricted list of indications stimulated immediate responses from a number of important organizations, including AACE. In the AACE response, the letter correctly pointed out the wide variety of conditions in which not diagnosing abnormally low vitamin D levels would put the patient at considerable risk for additional morbidity and mortality. I could not agree more with the AACE response.


But why should a website on patient safety talk about public policy? Because this policy will threaten patient safety, and it brings out the dilemma all of us must deal with each day. Health care must be affordable and costs must be carefully scrutinized. But building systems of care that are safer for our patients has real costs. It is not enough to be in favor of safer systems of care, we must also pay for them. We cannot run a hospital safely without numbers of well-trained nurses. We cannot expect to run a high-quality, safe, evidence-based outpatient care system when we take away key tools, such as 25-OH vitamin D testing, which is the best test for Vitamin D deficiency, and used every day in diagnosis and management of the disorders that need our care. The increased use of vitamin D assays today reflect a relatively recent increase in the understanding of the key importance of vitamin D levels in many conditions we treat, not just in bone disease, such as osteoporosis, but also in hyperthyroidism, diabetes, impaired glucose tolerance, in patients with severe muscle weakness and aching, certain cancers, and in some autoimmune disorders.


The story of osteoporosis care in the United States is particularly relevant to understand when it comes to the discussions surrounding 25-OH vitamin D testing. Osteoporosis is an extremely common condition that has been grossly undiagnosed and under-treated in our country for many years. Many of the hip and vertebral fractures that occur every year are preventable, but this proposed policy of restricting use of 25-OH vitamin D testing will be likely to increase the number of fractures, at a time where there already is an unacceptably high frequency of these fractures. In addition, not only will this policy worsen the outcomes of many patients with osteoporosis, but it will likely contribute to a large number of other problems for which the Medicare population is at risk. Since the wealthier patients will be more likely to afford to be tested for vitamin D deficiency when it is clearly needed, we are now introducing a serious threat mostly to our most vulnerable populations, a policy that both worsens patient safety and is inequitable and discriminatory. It will be the poorest citizens, particularly our minority patients, who will probably be most at risk for under-diagnosis and under-treatment of vitamin D deficiency related problems


The public may not be aware of what we are talking about. There are scientific studies to show that low 25-OH vitamin D levels are associated with an increased risk of muscle weakness, of falls, and an increased risk of hip and vertebral fractures when they do fall. But the effects of vitamin D deficiency are far more extensive than most people realize. For example, low vitamin D levels increase the risk of tuberculosis and hamper the ability to treat it. The risk for prostate, colon, non-Hodgkin’s lymphoma, melanoma, and breast cancers increases in the presence of vitamin D deficiency. Many medications, taken for other reasons, may cause low levels of the vitamin. Also, other diseases, such as hyperthyroidism and malabsorption, may also lower the levels of vitamin D. The deficiency of vitamin D may induce heart failure, worsen glucose intolerance, and increase the risk of many autoimmune diseases, including multiple sclerosis. Depressive symptoms worsen in the presence of vitamin D deficiency. In the United States, osteoporosis is very frequently associated with vitamin D deficiency, and in the presence of vitamin D lack, biphosphonate treatment is not enough. Appropriate calcium and vitamin D replacement must also be given and monitored if we are to reduce the risk of fragility fractures.


Policy makers understand that payment policies are very powerful tool for changing physician ordering behavior, and may be very useful in some settings, but they may not always realize that it is also a very blunt tool, which can cause great harm if improperly used. In this case, the scientific evidence, nicely summarized by M. Holick (1) in the New England Journal of Medicine in 2007, clearly shows the risk of vitamin D deficiency to all patients, both young and old, at all stages of life, but particularly in our ageing populations.


Five years ago, few physicians understood what we understand today about the ubiquity of vitamin D metabolites in our tissues and the central importance of those to our health. It may be that part of the issue is that the policy makers are not aware of the tremendous strides in the knowledge of vitamin D metabolism that we have made. But we know today that to restrict the ability of doctors to be able to perform this test on those who need it because CMS will not pay for it will have a dangerous effect on poorer and more vulnerable patients.


Surely the policy makers can do better and expand the list of approved uses of 25-OH vitamin D levels to one that is in keeping with the current scientific evidence.



    1. Holick MF. Vitamin D deficiency, N England J Med, 2007 July 19; 357(3): 266-281



Patient Safety and Racial and Ethnic Disparities


2009-02-27 08:43:18

By: Dr. Richard Hellman

The Third National Leadership Summit on Eliminating Racial and Ethnic Health Disparities was held on February 25-27, 2009, in National Harbor, Maryland. The conference has been widely anticipated and lived up to expectations. The presenters were many of the most well-known and well-respected individuals in the field of minority health and advocacy. Dr. Nancy Nielson represented the AMA, and her comments were very well received by the audience.


But the information presented that related to patient safety was particularly well done, and will be the focus of this editorial. Probably the most important fact was that there was much discussion by the speakers regarding the grievous medical errors that have resulted by failure to deal well with the linguistic, educational, and cultural barriers that so often occur with the care of ethnic minorities.


Most hospitals report that usually they are not made aware of a patient’s difficulty understanding or speaking English until after they have been admitted to the hospital. The languages are numerous; it is not only Spanish, but Vietnamese, Chinese, Sudanese, Polynesian dialects, and many other languages that may be spoken. The errors which are created when neither the physician nor the patient understands each other are both diagnostic and therapeutic in nature. Using a family member as a translator is often disastrous because they are not trained in medical terminology. Moreover, the family member really does not want to ask the questions, and the patient often does not want to tell their family member the answer. In this conference, the consensus was that, most often, family members find this kind of experience both demeaning and deeply upsetting. The best solution is not to use a family member, but instead a translator skilled in medical terminology.


Cultural difficulties are also an issue, and complicated by language barriers. The unsuspecting physician may not realize that apparent patient non-compliance may be based on strongly held beliefs that are cultural in nature, and these beliefs must be dealt with in order for good care to occur. The fatalistic view of illness in the Native-American culture is but one example of this phenomenon. The use of herbal remedies and other healing methods may also be accepted by the patient and their family and need to be dealt with in some way, for the herbs used may cause drug interactions that are clinically important.


Knowledge deficits among those from minority populations because of long-standing educational deficits are also common. Even if the physician’s patient can understand English, they may have no idea as to the context in which the physician’s explanation is framed. There is a regrettable tendency among a few physicians to consider the patients as stupid, or in some way unworthy of care, but prejudice has no place in healthcare. Fortunately, these kinds of events are becoming less frequent.


Another common error is to assume that the particular ethnic subgroup is uniform. In fact, a person of Peruvian origin is very different in many ways from their Spanish-speaking neighbor from Cuba. The same can be said for the European, Asian and African populations. The more we know about each person, the more useful information we will have to aid us in achieving an accurate and compassionate communication with the patient.


But most importantly, the ethnic minority patient is more apt to have a more delayed diagnosis of cancer, diabetes, or heart disease, and as a result, a more serious and complex problem. Their illness is more serious, to a significant degree because of the multiple barriers they face, limited access to care, lack of insurance, and more obstacles to achieving a healthy life style. It is then that the problem of communication tends to complicate the issue further, and make patient safety even more of a challenge. In the next months, we would like to share information that clinicians who deal with patients of the many ethnic and racial minority groups in the United States will find useful in achieving safer care.





The NICE-SUGAR Study on Intensive versus Conventional Glucose control-The Importance of Patient Safety in Achieving the Desired Outcomes


2009-03-25 15:56:43

By: Dr. Richard Hellman

The long-awaited NICE-SUGAR trial results have just been published on-line in the New England Journal of Medicine, together with an excellent editorial by Inzucchi and Siegel. The ADA and AACE have also released a joint statement. In their very thoughtful statement, they expressed a concern that providers of care will take the published results of the NICE-SUGAR study and conclude incorrectly that they can afford to be more complacent about uncontrolled hyperglycemia.


Their concern is justified. Some of the initial news commentary stated that the NICE-SUGAR showed that that intensive insulin therapy to achieve glucose control increases mortality in critically ill patients and may be dangerous. However, it is best to look carefully at what the actual data is.


The data provided by the NICE-SUGAR article shows that the odds ratio for death in the intensive control group was 1.14, with a 95% confidence interval of 1.02 to 1.28 and a P=0.02. The severe hypoglycemia rate (Glucose <=40mg/dl or 2.2mmol/L) was 6.8% in the intensive-control group and only 0.5% in the conventional group. On the surface it would seem that the results are unequivocal. But that would be an incorrect conclusion.


To begin with, the mean glucose level in the conventional-control group was 144mg/dl, as opposed to the intensive therapy group glucose level of 115mg/dl. The difference is not large, only a mean of 29mg%. The control in the conventional-control group was in the range of only relatively modest levels of hyperglycemia, much lower than the control groups in many other studies, which were often over 200mg%. . In addition, there were a significant number of patients, 74, whose vital status data was unknown, and a significant amount of reassignment of patients as noted in the paper. The 14% increase in death rate in the intensive group could be possibly explained by other factors not available in the analysis.


But a more compelling reason to be careful about the analysis is the fact that the NEJM article gives very little information about THE FREQUENCY THAT THE ALGORITHMS WERE DONE AS PLANNED. The algorithms used for the insulin infusions were constructed so the insulin infusion rate varied widely and the rate would change from 0 units/hr to 1, 2, and 4, units /hour in response to the glycemic levels noted. With this type of infusion algorithm, it is imperative to not let the interval in between the glucose determinations be excessively long, since otherwise the swings in glucose levels would be very rapid and large in amplitude.


In fact, in the report of their first 100 hypoglycemic events, the NICE-SUGAR study group, as reported on their ANZCA web-site, the most common adjudicated cause of hypoglycemia was clinician error, which was 37% of the group. They defined clinician error as either failure to follow the computerized treatment algorithm or inappropriately infrequent blood glucose monitoring. The next most common event, at 24 % of the adjudicated causes, was decreased nutritional intake. Each of these causes are a failure to follow the agreed upon plan.


So, is the problem of the failure of the NICE-SUAR study to show positive results in the intensive insulin therapy group that the target glucose level was set too low, or is it that the plan was good, but was carried out sub-optimally? While my opinion is that the reasons for the outcomes were complex, it is clear that in a sub-optimally conducted trial, even if the subjects were randomly assigned, a deviance from the plan of the trial weakens or limits the strength of the conclusions that should be drawn.


Yet, I think there is much about the NICE-SUGAR study that makes it an important addition to the literature. It suggests that in very ill patients, particularly patients on ventilators, and with marked severity of illness, that patients may be very vulnerable to hypoglycemia, and this much be scrupulously avoided. Also, the effects of early severe hypoglycemic episodes may be delayed, as some of the excess mortality was later, and may have represented subtle effects not initially noted in the ICU.


Also, the fact that a mean glucose level of 144mg% was associated with outcomes in an ICU that was not worse than those with lower glucose levels suggest that a target in this range may be safe.


But their own data on hypoglycemia strongly suggests that it is crucial to be sure that the algorithm is being carried out correctly for optimal results. And data from both Keukoven, 2008, Blaha, 2009, and Van Den Berghe, 2008, all point out that the training of a nursing team who is in change of actually overseeing the insulin infusions is crucial. There is a need to provide adequate numbers of trained personnel, and both education and continuing oversight. Quality improvement in this setting depends upon reviewing the work product and improving what is less than optimal. . In the editorial by Inzucchi and Siegel, they correctly point out that in many hospitals, a seamless and automatic use of insulin infusions has developed. These efforts have already resulted in a small but growing literature on the successful use of insulin infusions. More data will certainly be welcomed, for the NICE-SUGAR study is important, but far from definitive.


    1. The NICE-SUGAR Investigators, Intensive versus Conventional Glucose Control in Critically Ill Patients, NEJM, 360;13 March 26, 2009, pgs 1283-1297
    2. Inzucchi, SE, and Siegel, M., Glucose Control in the ICU-How Tight is Tight?, NEMJ, 360;13, March 26, 2009, pgs 1346-9
    3. Joint Statement of the ADA and AACE on the NICE-sugar study, Press Release, March 24, 2009, on




The NICE-SUGAR trial and the implications for clinical practice settings – Can We Generalize the Findings?


2009-04-06 16:26:12

By: Dr. Richard Hellman

The NICE-SUGAR findings have been the source of much discussion, and we have devoted several earlier editorials to this subject. But there is great interest of both our readership as well as the larger medical community, and more discussion is warranted. We will begin today with a re-examination of the paper itself.


The strength of any multi-center RCT such as the NICE-SUGAR trial lies in the large number of subjects. The statistical power of the large number of subjects with similar outcomes lowers the probability that the outcome could have occurred by chance. But such a design does not protect against either excessive, unmeasured variation in the use of the protocol, or if unmeasured events occur which have a significant influence on the measured outcome. Post-randomization events are a particular problem for many RCT’s because they often are assumed to be minimized in importance by the randomization. However, this may not at all be the case.


The NICE-SUGAR data that was provided on their website, a subject of my February 23rd editorial, shows that there was significant deviation from their protocol. Of the first 100 cases of severe hypoglycemia, all adverse events, 39% were adjudicated to be due to clinician error, and 24% to under-nutrition, both deviations from the protocol. But the effects of deviation from the protocol can not be assumed to have an equal impact in both arms of the study. For example, if a patient was in the intensive therapy group, it was likely that the glucose levels were lower to begin with, and a hypoglycemic event due to clinician error may well have resulted in even lower glucose levels than in the comparison group. We do not have data provided to be sure, but this is a reasonable assumption. Unfortunately, the investigators were not forthcoming in their release of the data as to how frequently glycemic levels were performed, not just how often the protocol required, but how often they were actually done. Again, if hypoglycemia increased mortality in some vulnerable patients, it is likely that deviations from protocol would have a greater impact with lower or more prolonged severe hypoglycemic levels.


One of the curious results of the NICE-SUGAR study was that at day 28, there was not a statistically significant difference in mortality between the intensive group and the conventional treatment group, although there were more deaths in the intensive therapy group, 22.3% vs, 20.7%, p=0.17. The differences noted that were statistically significant were at 90 days, where the intensive therapy group mortality was 27.5% vs. 24.9% with a p=0.02. The key question is: What happened in the 84 days after these patients were in the ICU (almost of whom were initially on a ventilator). There is not a good deal of data in the NEJM article to help us in this regard. This is unfortunate, since post-randomization events here may be significant, and little data is provided to show us what did happen. What we do know is that a significant number of patients, some by personal or family choice, others by physician choice, later withdrew from the study. These numbers were large enough to have also blurred the results if the choices for withdrawal were not randomly decided upon.


Does the study have generalized applicability? Does it supersede or invalidate the earlier Van Den Berghe data? No, the lack of detail as to how the protocols were performed in NICE-SUGAR, how often glycemic levels were actually checked, as well as lack of data on post-ICU, post insulin infusion events, make us hesitant to jump to that conclusion. Also, the missing data from patients who withdrew from the study also makes the statistical power of the study less certain than we would like. Additionally, the overlap between the glycemic levels of the two groups, separated by a mean glucose level of only 29mg% also suggests to us that the 90-day mortality data may have other causes than just the differences in glycemic levels in the initial ICU stay.


In a way, however, despite justified criticism of the NICE-SUGAR study, this study did show us that attention to how a study is performed, particularly in complex environments such as a hospital, is crucial for its success. Also, the study results suggest that both short and long-term effects of hypoglycemia may be significant and should be avoided.


Short-term data shows that hypoglycemia greatly increases the risk of ventricular arrhythmias, in part due to the prolongation of the QT interval (Cryer et al, 2009) , a high risk for those at risk for ventricular arrhythmias, such as those patients with myocardial ischemia or ventricular dilatation. But also, there is other evidence that patients with CNS disease may be particularly prone to severe hypoglycemia during insulin infusions. It is probable that there was a subset of patients in the NICE-SUGAR study that were at particular risk for mortality due to hypoglycemia in this very ill population. More than 90% of the patients in NICE-SUGAR were on ventilators during the initial phase of the study. In the longer term, there is other data to show that triggering of inflammatory cytokines occurs during hypoglycemia, and this may also be a risk for poorer outcomes at 90 days, but we do not know if this was an important factor in this trial.


Severe hypoglycemia may well be deleterious to vulnerable ICU patients. In order to protect patients from severe hypoglycemia, while avoiding severe hyperglycemia, which is well-documented to be a severe risk factor for a poor outcome, it is best when instituting a protocol for intensive insulin therapy to pay attention to the support systems, the nurses and other staff, as well as the education received by all those who are involved in implementing the insulin infusion, which is the mainstay of the ICU care. We need to check the glycemic and clinical outcomes in our centers, re-evaluate work flows and optimize them in an iterative fashion so we can be sure that we are achieving what we wanted to achieve.


Setting the bar for glycemic control too high, however, is a poor choice, for even setting a glycemic level of 180mg% is no assurance against hypoglycemia if the infrastructure of care is lacking, and the dangers of significant hyperglycemia in the vulnerable patients is very great.


It is likely that many hospitals can initially only safely achieve targets of 140-180mg% in an ICU setting without unacceptable hypoglycemia. But other centers have been safely achieving lower targets, in the 110-140mg% range. It is not at all clear whether it is safe to aim at targets lower than 110mg%, particularly if point-of-care glucose instrumentation is not sufficiently accurate, a common problem today, and one discussed in my editorial of February 25th on bedside glucose monitoring. For each center, however, it will be important to frequently measure their own glycemic and clinical outcomes, and see whether their improvements in their infrastructure are sufficient to achieve the results they want.




Jerome Groopman’s article in Wall Street Journal: “Why ‘Quality’ Care is Dangerous”. Is he right – or wrong? Or both?


2009-04-09 14:41:51

By: Dr. Richard Hellman

On April 8th, the Wall Street Journal published an op-ed article by Jerome Groopman and Pamela Hartzband. The article was sharply critical of a reported effort of Medicare to tie pay-for-performance incentives to quality metrics that they have designed. He points out that the state of Massachusetts is also using “quality metrics” and has developed an elaborate and punitive system to force doctors and hospitals to conform to their standards.


In the article he gives several examples where recent published data are providing evidence that the metrics used by Massachusetts may be wrong. Two of the examples he uses are the NICE-SUGAR trial, reported this March in the NEJM, and the ACCORD trial, published in the NEJM in June 2008.


While Groopman and Hartzband are right regarding the ill-conceived plan of both Medicare and Massachusetts to institute pay-for-performance initiatives that in both cases run beyond the clinical evidence, they are wrong in their interpretation of the two trials they used as evidence for their position. In both of the trials, the poorer than expected outcomes may have been due to factors not anticipated by the study investigators, unmeasured events that altered clinical outcomes. In the case of the ACCORD trial, a trial that ran counter to the results of both the ADVANCE trial and the VADT trial, there is evidence that undiagnosed hypoglycemia may have been more frequent than they had observed. There is also suspicion, based on the comments of the investigators, that the complexity and rigidity of the protocol may have been very difficult for many of their patients and may have resulted in patient errors and deviation from their plan.


In the case of the NICE-SUGAR trial, the deviations from the protocol were significant, as we have previously reviewed, and the lack of data regarding post-randomization events makes the final results less likely to be widely applicable.


While Groopman is right that it is unreasonable to do “corrective” counseling for a doctor whose patient has glucose levels that are higher than the original targets suggested by Van Den Berghe in the Surgical ICU study in 2001, he is wrong that the two studies discredit the notion that intensive control of glycemia does not have merit. It does, in an appropriate setting, with appropriate systemic support, and as long as control of glycemia is not resulting in unacceptable levels of hypoglycemia.


However, I think Groopman and Hartzband are more right than wrong, for the great danger of the quality movement in medicine is that the process is at risk of being politicized, and politics is running ahead of the evidence, focusing on what we would like to accomplish, instead of what we can do that is scientifically grounded. His discussion of the current misapplication of quality metrics in the state of Massachusetts is probably one of the best examples of politicians run amok – choosing to punish or reward doctors based on flimsy scientific basis.


On the one hand, the work of Carolyn Clancy, the administrator of the Agency for Health Care Research and Quality (AHRQ), is a model of how the scientific method can be applied to quality improvement in a positive way. The AHRQ website has evidence of comparative effectiveness research that is backed by evidence. For example, the current data showing the pluses and minuses of various fixed combination insulin preparations are presented together with the appropriate supporting citations. In an editorial that she wrote for the public, she stated, “In health care, one size does not fit all, and treatments that are good for one patient aren’t necessarily good, or even safe, for you”. Her comments are a far cry from others’ attempts to use claims-based data, without any risk stratification adjustment, to make judgments on the quality of care.


Mark Chassin, the President of the Joint Commission, in a recent statement and pod-cast which can be found on the AHRQ M&M website, made the point that claims data-bases do not produce as highly valid quality measures as most of the clinically based core measures. He also does not believe that information technology is the solution to all our quality and safety issues. He stated, “Computers do not make us less stupid, they make us stupid faster”. His point is that a poorly designed IT solution, for example a Computer Physician Order Entry system, can do a great deal of harm. I think he is keenly aware of the dangers of assuming that technology alone can improve the processes of care that are so broken in many of our health centers.


I think there is more than enough data to tell us that quality of health care and patient safety issues are a world-wide problem in both inpatient and outpatient settings. The settings are different, the focus and resources may be very different, but we see commonality wherever health care is being provided.


To improve the quality of care beyond where we are at present, we need to be both more humble and more scientifically grounded. We may want to improve one hundred different clinical processes of care, but if we have data to clearly indicate the best approach in only ten processes, then we should focus on the ten processes we know have the evidence to support what we intend to do.


We cannot sacrifice scientific rigor because the public demands we do something, even if it is wrong. If we know that co-morbid or demographic factors play a significant role in a measurable outcome, then when we measure the outcome, we need to use the co-morbid or demographic data so we can fairly assign responsibility to those who are involved in the care. To reward or punish health care providers for things for which they cannot control will only cause cynicism and resistance, and, ultimately, a public outcry and failure of the effort to improve quality.


The middle ground is probably to encourage ongoing scrutiny of important clinical data for quality improvement, and to link reporting of clinical data to quality improvement efforts. If we just focus on reporting, we are missing the whole point of what we are trying to do, to improve quality of care and safety as measured by clinically important outcomes. It is not easy to change processes of care, to create systems that capture errors before they injure patients. It is even harder to develop the discipline to check your own work, to admit that you may be at times prone to diagnostic errors, and to work with systems that will make the possibility of patient injury remote.


The clinical community needs to embrace quality improvement, not just by words, but by deeds. But it is much harder to force change by using a carrot and stick approach, and the pay-for-performance efforts at present, are, for the most part, not scientifically sound. We need to redouble our efforts to be sure that the health policy community and those in Washington understand why we need to make scientific validity a central part of our efforts to improve safety and quality of care.




Dr. Hellman Comments on PTU Editorial


2009-04-22 14:12:07

By: Dr. Richard Hellman

The AACE Patient Safety Exchange is happy to publish an editorial by three distinguished endocrinologists regarding their well-documented and carefully constructed opinion on the safe use of antithyroid medication in pediatric populations. In their editorial, they also discuss the safety of Propylthiouracil (PTU) in both pregnancy and in non-pregnant adults with hyperthyroidism.


We welcome further input and discussion by our readership. The thoughtful comments of Dr Rivkees and his colleagues deserve serious thought and analysis. I want to thank them for their submission, and I hope the discussion will be lively.


Richard Hellman MD, FACP, FACE


AACE Patient Safety Exchange


Propylthiouracil (PTU) Hepatotoxicity and Graves' Disease Therapy





A recent report in JAMA on the association of hypoglycemia and dementia in older patients with type 2 diabetes – Is the link causal? And if so, in what direction is the causation?


2009-05-11 14:05:15

By: Dr. Richard Hellman

A recent article in JAMA by Whitmer et al reports on an association between the diagnosis of severe hypoglycemia and a subsequent diagnosis of dementia1. The authors conclude that their data makes it unlikely that the reported association is due to dementia increasing the likelihood of hypoglycemia, but rather that early hypoglycemia in this population increases the later risk of dementia. They suggested caution on therapy in older patients with type 2 diabetes. Further, they point to the ACCORD trial as reporting related findings.


Their longitudinal cohort study was based on a Kaiser Permanente Database, and was conducted using data from a period of 1980-2007 on 16,667 patients with a mean age of 65 years and type 2 diabetes, who were members of an integrated health care system in northern California.


Their results showed that of the 1415 patients (8.8%) with one documented episode of severe hypoglycemia, dementia was subsequently diagnosed in 250 patients. They concluded that fully adjusted hazard ratios for one episode of hypoglycemia was (HR, 1.26; 95% confidence interval [CI], 1.10-1.49); 2 episodes (HR, 1.80; 95% CI, 1.37-2.36); and 3 or more episodes (HR, 1.94; 95% CI, 1.42-2.64). They found that the results were similar but slightly more positive when looking at ER admissions for hypoglycemia.


Unfortunately, this is a flawed and misleading study. Because it was in large part dependent upon incomplete data sources derived from hospital and outpatient and ER admissions, there is little question that the identified episodes of hypoglycemia were only a fraction of the total episodes of hypoglycemia experienced by the patients. Also, there was no baseline systematic evaluation of cognitive function of the cohort, and the standard for diagnosis of dementia was not uniformly applied. The same can be said for the diagnosis of hypoglycemia.


Additionally, the group with an episode of hypoglycemia was older at the time of the survey, had a longer duration of diabetes, and had a higher co-morbid index of heart disease, hypertension, previous stroke, and end stage renal disease. Although the statistics on the differences in duration of diabetes was not given, all of the others mentioned were statistically significant.


It should be expected that there would be a higher incidence of cognitive dysfunction in this group, as each of these are accepted risks for cognitive dysfunction in adults. On the other hand, there is no information given on the social support available for the patients with hypoglycemia, whether they lived alone, or the frequency of glucose self-monitoring, the lack of which in insulin users is associated with an increased risk for injury (D Cox, EASD Rome 2008). Neither is there information about depression in diabetes which increases the risk for hypoglycemia and cognitive dysfunction. The data does show an association, but it does not answer the question of causation.


One can easily posit a quite different explanation for the observed association of hypoglycemia and dementia. This is as follows:


Severe hypoglycemia may be a marker for patients at higher risk for dementia and in fact may be more commonly seen in patients who already have mild-to-moderate cognitive dysfunction. Those patients have poorer ability to protect themselves and provide optimal self-care, and may be more prone to both hyper and hypoglycemia. The higher glucose levels and increased glycemic variability will increase their risk for earlier and more severe complications, some of which, in turn, will increase the risk for more hypoglycemia as well as a poorer overall clinical outcome. It is no surprise that such patients have increased death rates, a common feature of patients with increasing cognitive dysfunction.


While it is true that severe brain injury can follow an episode of prolonged hypoglycemia, particularly after seizure activity, the data from the EDIC trial suggests that severe hypoglycemic episodes may not necessarily lead to long term dementia, and in fact, there is a better correlation with long term A1C values with respect to cognitive dysfunction.


But the case in older adults is not settled and the study of Whitmer et al is not the definitive answer at all. Rather, it highlights the need for us to study this vulnerable population, patients 60 years of age or greater with diabetes and co-morbid vasculopathy, and to carefully examine their baseline status regarding both depression and cognitive function. There are a variety of easy-to-use validated tools for rapid evaluation of both depression and dementia. Both the Mini-Mental Status Exam and the Beck Depression Inventory are examples of the many good tools available.


At the end of the day, by proper identification of the potential vulnerabilities of each of our at-risk populations, we can improve our ability to tailor the diabetes therapy to the actual needs of the patient and make diabetes care not only more effective, but safer as well.



    1. Whitmer RA, Karter AJ, et al. Hypoglycemic Episodes and Risk of Dementia in Older Patients With Type 2 Diabetes Mellitus. JAMA, April 15, 2009; 301: 1565-1572.





The AACE/ADA Consensus Statement on Inpatient Glycemic Control and Patient Safety – What key points were made and what was still left unsaid?


2009-05-29 14:31:41

By: Dr. Richard Hellman

On May 8th, 2009, the AACE/ADA Consensus Statement on Inpatient Glycemic Control was published on-line, simultaneously in both Diabetes Care and Endocrine Practice. The ten-member panel of experts, co-chaired by Etie Moghissi and Mary Korytkowski, worked together since late 2008 to put together this document. I was honored to be one of the ten members who wrote the paper. My responsibility was to focus on patient safety, systems that promote safety, and also on glucose meters, their deficiencies in the areas of accuracy and precision, and how those deficiencies impact the safety of the patients.


It will be no surprise to all who read the document carefully that patient safety was treated both carefully and in great detail. Perhaps because so many of the outstanding multi-disciplinary panel are deeply involved with the day-to-day problems in safely achieving inpatient glycemic control, great care was made to assure that the document is very balanced, and to not overstate the conclusions of the authors. For example, the statement that the recommended target glycemic level should be 140-180mg% is paired with the statement that "greater benefit may be realized at the lower end of this range". It is then followed by the statement that lower glucose targets may be appropriate in selected patients, but targets under 110mg/dl are not recommended.


The authors' reasoning was that the evidence from numerous recent randomized clinical studies showed increased harm at the lowest targets. Even though there was earlier evidence, both from the Van Den Berghe trials and from other non-randomized trials, of the efficacy and safety of lower targets, those studies were relatively few in number, and in the face of conflicting studies, the writing panel agreed that it was better to be conservative in the recommended targets.


It is noteworthy that both in the article and in the discussions that led to it, many of the panelists, including myself, were concerned that people might look at the target levels noted in the paper and conclude that glucose levels of 180mg% or greater might be acceptable. In fact, there is very strong evidence that such levels are harmful to inpatients with hyperglycemia and lead to adverse outcomes in both mortality and morbidity. Another important conclusion, both stated and implied, that can be fairly drawn from the paper was that the risk/benefit ratio for any individual patient is often not the same and individualization of therapy is important to provide the best outcomes.


In the consensus statement we pointed out the many systemic problems that may present barriers to safer clinical care and how to look for them. There was a brief discussion of the limitation of point-of-care glucose meters, both regarding their inaccuracy in special situations, and the more general issues that currently limit the usefulness of these devices. At present, many hospitals take for granted that the glucose meters they use are giving accurate and reproducible results. Their hospital personnel are basing insulin doses upon the reported glucose levels obtained from these meters. However, the reality is that many of the glucose meters in common use are affected not only by interfering substances that differ between types of glucose meters, but also by other factors such as the hematocrit of the patient. One of the underlying problems for all of the meters is that the standards for the accuracy and precision of the meters, which are set by the FDA, are so low that relatively large and potentially dangerous errors in insulin administration can and do result. Additionally, many hospitals do not either train or supervise how their personnel use these glucose meters, which need to be used in a very consistent way in order to obtain optimal performance. Errors due to faulty operator technique add to the inherent problems of inaccuracy of many of the meters.


An important method that is crucial to maintaining a high quality of glycemic control is to properly use glucose metrics. The consensus statement contains a recommendation that glucose metrics be kept to monitor the performance of glycemic control. These metrics are used for evidence-driven decisions about glycemic control. Keeping glucose metrics regarding the actual day-to-day performance of the hospital staff in glucose control is key to an effective and safe program of glycemic control.


One of the strongest recommendations of the writing panel was regarding sliding scale insulin. Sliding scale insulin as monotherapy is both ineffective in most patients and potentially dangerous in Type 1 diabetes. The paper provided the evidence to support this important conclusion.

As pleased as I was with the high quality of the consensus statement, and as grateful as I was to the expertise of all those who participated, I was keenly aware that for many of us, the consensus statement was the result of a decision to focus on what we could agree upon. Each of the authors could also articulate other areas of importance that were not fully included in the statement.


So, while I agree with what we wrote, I would like to talk about important areas in the field of patient safety that were not included in the very focused document. The first point is regarding the NICE-SUGAR study. Although not included in the published article, the ANCZA website had posted preliminary data from the NICE-SUGAR study that showed how often hypoglycemia was the result of a deviation from their own protocol, most often due to clinician error. In addition, even one of the reviewers of their published article in NEJM was unable to obtain from the authors the data on how often the glucose monitoring was actually done in the NICE-SUGAR study. All we know is how often the glucose monitoring was intended to be performed, and this actual difference could be crucial.


If the protocol for glycemic control is excellent, but the protocol is not performed as planned, the results should not be expected to be optimal. The systems that we talk about in patient safety discussions most often are the work-flows, equipment, training, teamwork, coordination, and metrics needed to make the protocol work as planned, every time. One of the problems with the negative results reported by some of the authors in both the reported randomized trials and in the meta-analyses, was that the negative results may have been due at least in part to the fact that investigators did not pay as much attention to developing the systems of care as they needed to to get the results they desired.


In fact, it was clear that using sophisticated and effective systems of care, several of the authors of the AACE/ADA consensus statement were able obtain excellent glycemic control with targets in the 110-140mg% range without overly-frequent severe hypoglycemia. The key appears to be setting appropriate glycemic goals that can be achieved safely and consistently. These goals are most often achieved by a multi-disciplinary approach with careful attention to making the system of care sufficiently robust in order to attain the glycemic goals selected. It may be best for an inpatient center to set a goal of 140-180mg% and as they get better in their coordination of their efforts at glycemic control, lower their glycemic target towards 110mg%, while making sure that they keep carefully studied glucose metrics so they can be sure that they are doing what they set out to do without causing harm to the patients in the process.


What we also did not discuss in the consensus statement in great detail is the very large problem of the current accuracy and precision of glucose meters and how important it is to be sure that the method used to obtain timely bedside glucose determinations that work well. There was a statement made in the text of the statement that the FDA standard of accuracy allows a ±20% error for glucose meters, and "questions have been raised about the appropriateness of this criterion". That is, in my opinion, a very large understatement. In fact, many experts agree that improving the accuracy of the glucose meters used in point-of-care testing is crucial, since we make insulin dosage choices based on these numbers. The phrase, "garbage in, garbage out", seems particularly relevant here.


Also, largely because of space considerations, there was not a discussion in the paper about Medicare "Never Events", or preventable diagnostic conditions that have developed in a hospital setting. In the area of glycemic control, the 2007 Medicare data lists hypoglycemia as a far less common diagnosis than either diabetic ketoacidosis or hyperosmolar non-ketotic hyperglycemia. I have been surprised by the relatively large number of hospital acquired DKA and HHNK events reported by Medicare, events that, we would all agree, should never be allowed to occur during a hospital stay. Although this subject was not discussed in the consensus statement, it will be the subject of subsequent editorials on this safety website.


Nearly all of the panelists agreed that more studies are needed comparing different methods, including insulin infusions, in the non-critical areas of hospitals. It is clear that randomized trials in this subject area are far fewer than needed. In recent years, the National Institute of Health has not funded such trials, but if society wants safety in glycemic inpatient care as a feature of the health care system, then society needs to fund appropriate studies. The Agency for Health Care Research and Quality (AHRQ) probably will be the appropriate site of some of the funding. Other not-for-profit agencies interested in patient safety should seriously consider funding more studies in the non-critical areas of hospitals so we can have the appropriate evidence base we need for extending the umbrella of good glycemic control for all hospital inpatients.


In summary, the new AACE/ADA Consensus statement on Inpatient Glycemic Control should be required reading for all clinicians who care for hyperglycemic patients in hospital settings. It is an excellent starting point for further discussions, and we have tried to begin the discussion on this page today. We welcome your comments.



    1. American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control.
    2. Moghissi ES, Korytkowski MT, DiNardo M, Einhorn D, Hellman R, Hirsch IB, Inzucchi SE, Ismail-Beigi F, Kirkman MS, Umpierrez GE; American Association of Clinical Endocrinologists; American Diabetes Association. Diabetes Care. 2009 Jun;32(6):1119-31 (co-published in Endocrine Practice, Vol 15, No 4, May/June 2009, pp1-17)






The FDA and AACE – Discussions on a new medication, Cycloset.


2009-05-29 14:42:58

By: Dr. Richard Hellman

Last week, the FDA engaged in a discussion with some of the AACE leadership about a new hypoglycemic agent, Cycloset, a quick-release low-dose formulation of bromocriptine mesylate, a dopamine agonist, which is thought to act centrally to reduce resistance to insulin-mediated suppression of hepatic glucose output and tissue glucose disposal. The FDA Division of Metabolism and Endocrinology Drug Products, which is directed by Mary Parks, MD, provided the information. Dr. Parks and her colleagues were participants, as were others from the FDA, joined by key members of the AACE leadership.


The drug had been just approved for use in the United States, and the discussion allowed the FDA to directly receive input from thoughtful clinicians about the agent, and for AACE to have an opportunity to hear how the FDA scientists think about drug safety and how they evaluate the relative risks the drug poses for patients.


A recent randomized clinical trial of 12-month duration with over 3000 patients had just been reported that failed to find any evidence of significantly increased cardiac risk, but the drug is not recommended for any woman who is postpartum, because of the seizure risk, or any patient who has orthostatic hypotension because of their increased sensitivity to dopamine agonists and the risk of more severe hypotension. It is likely that with more widespread use, more information about the safety, tolerability and efficacy of this medication at the recommended dosage will become available. Cycloset is thought to be a drug of relatively modest potency that is likely to be used with other agents, such as insulin secretagogues, metformin, and exogenous insulin.


While the basic information that was discussed on the conference call is already in the public domain, the meeting appeared to be a very important step forward for the FDA. Their intent is to continue reaching out to specialty societies, to both share information and listen to responses from clinicians.


It also seems clear that the FDA would very much appreciate increased voluntary reporting activities of the clinicians. The FDA appears acutely aware that many problems of great importance but relatively low frequency will first be noted by clinicians in the community, and they hope that we will send more information to them about associated adverse events, which they can then evaluate and study in more detail. Some societies, such as the American College of Cardiology, collect clinical data on drug safety and efficacy, as well as quality, on behalf of their members, and use this data to educate their members and improve clinical care.


This meeting, however, was noteworthy because it was the first collaboration between AACE and the FDA, with a goal to provide timely information to our membership regarding a new use for an old medication. We hope this is the first of many such discussions as more new medications appear on the scene.




The FDA Med-Watch Alert on Stolen Levemir – Is this just one example of a much bigger patient safety problem?


2009-06-17 16:22:39

By: Dr. Richard Hellman

On June 13th, the FDA released a Public Health Advisory concerning Levemir, the long-acting insulin made by Novo Nordisk. As many as 129,000 vials were stolen, and many of those may have found their way into the marketplace and may be in the possession of your patients. The lot numbers are XZF0036, XZF0037, and XZF0038.


The problem for the patient is that the insulin vials may have been mishandled after they were stolen. Exposure to extremes of heat, cold, or excessive vibration can easily degrade the insulin and decrease the efficacy of the affected vial. The patient may not see evidence of cloudiness or discoloration of the vial or its contents, and the only clue may be the fact that the insulin does not lower the blood glucose as well as it did before.


However, many of you may be thinking, "I heard this news from my specialty society late on Saturday or Sunday, so why are we hearing about it again?" It is because there are a few facts that are important to underscore.


The first, and perhaps the most important, is that the speed with which the FDA reached out to the medical specialties is clearly due to a new policy direction of the agency. They are making great efforts to engage the societies as well as make the news available to all in a very timely manner. In fact, the surest way to get the FDA alerts and warnings is to subscribe to their services, so they will deliver their alerts to you as soon as possible. The link is: Their only requirement is that you provide them with your email address. Then you can choose whichever alerts that interest you the most. You can also have your practice administrator sign up, so alerts that will affect your patient population can be responded to as quickly as possible, regardless of whether you are at the office or away.


Also, it is fair to ask whether it is time for the manufacturers to consider whether all containers of insulin vials should have an indicator that is temperature sensitive in order to show if the individual vial has been exposed to excessive heat or cold. When shipping large or valuable containers, many industries will put such temperature indicators to indicate when a temperature sensitive product has been exposed to an unacceptable temperature. Insulin is often shipped in this way, but indicators are not provided at the individual vial level, presumably because of cost.


But how often has a patient inadvertently placed their insulin in a place where it was exposed to heat and not realized it until too late? There is a need to consider whether we should utilize the technology already available to add this kind of indicator to the vials. It is important that the user of the insulin know that no one in the long chain of those who handled the insulin before they use it, including themselves, allowed the insulin to be exposed to temperatures that would degrade it. If adding such indicators can be made economically feasible, and I think it can be done, it would protect the insulin-requiring patient more than they are at present. The example today, the major theft of Levemir insulin, is just one example of the potential dangers when insulin integrity is compromised.




Patient Safety, Insulin Pumps, and Cognitive Dysfunction


2009-06-22 08:36:35

By: Dr. Richard Hellman

Earlier today, I consulted on a patient in the hospital who was there because of recurrent hypoglycemia. It was an interesting problem. The patient was an 85-year-old man, with a 39-year history of insulin dependent diabetes, who had been on an external insulin pump for twenty-five years. During the past two years, he had been hospitalized four times, once for pneumonia, once for diabetic ketoacidosis, and twice for hypoglycemia. During this time, his estimated glomerular filtration rate dropped from 33ml/minute to 18ml/minute, and his A1C had gradually fallen at the same time from a mean of 7.3% to 6.8%.


On the surface it would seem straightforward to conclude that, as the renal clearance of insulin decreased, and the capability of the kidney for gluconeogenesis decreased, the patient's risk for hypoglycemia would increase and may have been responsible for the problem. There is no question, in fact, that these factors played a role in increasing this patient's risk for hypoglycemia. In general, not just chronic renal failure, but liver dysfunction and congestive heart failure may predispose patients to fasting hypoglycemia, and unsuspected endocrine disorders such as hypothyroidism, or primary or secondary adrenal insufficiency, should also be considered as potential causes of fasting hypoglycemia due to increased insulin sensitivity.


But the point of the editorial is not one of the well-studied causes listed above. The patient, once a very alert gentleman, could not remember any of the five basal insulin rates his pump was programmed to deliver. Neither could he explain the reasoning behind the doses of insulin he uses for treatment of hyperglycemia. At least one of his decisions to treat hyperglycemia appeared to be illogical and was followed by hypoglycemia. He was vague in his recollections regarding other historical events as well. It was clear that he had suffered considerable cognitive decline, a common finding in patients with diabetes over sixty-five years of age, and much more common as renal function declines to moderately severe levels. Also, he had been very depressed since the death of his wife of many years, from cancer, earlier this year. As for his physical condition, using the National Kidney Foundation terminology, he had gone from Stage III renal failure, to between 15-30ml/min creatinine clearance, Stage IV, the most severe stage of renal failure that is not routinely treated by dialysis.


It is likely that his increased frequency of hypoglycemia was related not just to the altered insulin kinetics, but to the cognitive dysfunction, which resulted in poorer decisions and frequent hypoglycemia. His depression almost certainly was a complicating factor as well. Further complexity is added by the fact that although the endocrinologist in the ambulatory care setting had been trying to encourage the patient to consider using multiple daily doses of subcutaneous insulin instead, the patient had resisted the physician's efforts to discourage insulin pump use for him at this time.


The data available on cognitive decline shows that cognitive dysfunction is a common feature in type 2 and type 1 diabetes, and, depending upon the tools used to measure the cognitive dysfunction, many patients will show evidence of this finding if carefully tested. A careful analysis of type 1 patients without dementia will show a subtle cognitive decline in many patients, even in childhood and adolescence, but these are relatively minor changes and are usually not clinically significant, unlike the problems later in life that may lead to dementia. In all diabetic patients, some intermittent problems, such as hypoglycemia or ketosis in ketosis-prone patients, can cause short-lived, but important, dysfunction.


More important is the data that shows the close relationship between cognitive decline and the risk factors of: increased age, duration of diabetes, degree of control of glycemia long-term, hypertension, other forms of cardiovascular disease, and depression. Depression may confer up to a 2.5-fold increase in risk of cognitive decline (Katon 2009).


Should we not use the available tools to measure cognitive function prior to putting patients on a complex device such as an insulin pump? Should we do formal testing, as the French do, to ascertain whether the patient has retained what they have been taught related to the use of an insulin pump? Should we retest the patients periodically, especially after they reach the age of 60 or greater? It is likely that this patient needs a careful reassessment prior to restarting the insulin pump. But just as patients who have diabetes and drive automobiles fight fiercely to keep their license, so do some patients regarding their insulin pump.


Although there are insulin pumps that are better designed than others, there is no Consumer Reports information to guide the unwary patient or doctor. Many patients are placed on pumps without a careful evaluation by the physician who will be supervising their pump care. In many cases, neither the patient nor the physician is aware of more than a small fraction of the capabilities of the insulin pump, and neither is particularly well informed as to the full menu of functions and the sequences of activation of commands. Often, the only contact with a professional knowledgeable about the pump the patient encounters is the pump trainer of the of the pump manufacturer, a paid employee who would be unlikely to give the patient, the new pump user, a completely unbiased opinion regarding the relative value, comparative features and safety of their pump for that patient. We need more access to just-in-time comparative information about the pumps so the process of care can be more transparent, and the physicians and other staff who perform this task need to be reimbursed for their time and expertise.


This particular case illustrates a key issue in patient safety, the need to reevaluate our patients regularly, because as disease states change in severity, and the patient's physical, mental and emotional states change, a most appropriate therapy may become dangerous, as it has here, due to combination of each of those factors.




Patient Safety and Bilingualism


2009-06-23 13:40:32

By: Dr. Richard Hellman

At this year's annual meeting of the American Diabetes Association, several sessions were devoted to the barriers to care that can occur when a patient who does not speak English as their primary language is taken care of by a provider who speaks only English.


The problems are well known, but poorly understood. For instance, when a patient for whom English is not their primary language communicates, do you have an idea of the level of their proficiency? Some patients may be able to converse in English, but their comprehension of the written word is much poorer. The vernacular, the common expressions we take for granted, may be a mystery to the patient, and may vary dependent upon their age and cultural heritage.


When using a translator, differences in culture may add a layer of confusion, since a translator who is from Peru will not necessarily be helpful with a Columbian or Mexican patient, because, although they both speak Spanish, the language is used significantly differently, so that confusion is not uncommon. Also, the prejudices of the translator may affect both how the questions are asked and the translation of the reply.


When it comes to cognitive testing, few of the tools used are validated across languages and different cultures. Put simply, the tools we use to check for dementia are likely to give misleading information when tested on a patient who speaks another primary language. Given the high degree of mild cognitive dysfunction in diabetics who have cardiovascular or renal disease, or are over 65 years of age, better, validated tools are urgently needed. At present, the definitive evaluation of cognitive testing is a very time-consuming, laborious endeavor, and this limits the large scale use to research studies.


However, this may be changing soon. There are several attempts to use on-line cognitive function testing, and at least several of the tests do not depend upon language proficiency, since response time, memory, spatial relations and computations are not dependent upon the subject's language skills. While it is premature to adopt any of the current on-line methods, they are most promising, and may provide a new way in the future to assess the functional cognitive status of a patient, both at baseline and serially, so we can see how their function changes and use it to guide when complex therapies are safe for a patient to use.


Fortunately, there are many organizations that are dealing with racial and ethnic barriers and are urging the development of tools that will be useful when confronted with the need to care for patients for whom English is not their primary language. One of the newest organizations is the National Council of Asian and Pacific Islander Physicians, led by Dr. Ho Tran, which is working on a national level to bring resources to the care of the many ethnic groups that they represent. They join the National Medical Association and many other groups that hope to make a difference in these areas.




Patient Safety Issues in the U.S – The dangers of an overly narrow focus


2009-06-24 11:41:18

By: Dr. Richard Hellman

One of the advantages of inpatient consultations in endocrinology and diabetes is that it allows an opportunity to see, in situ, the results of decisions made by others as they happen. One such example presented itself to me this morning.


The patient was a young man, age 31, who presented to the hospital with mild diabetic ketoacidosis. He had glucose of 1032mg% and mild acidosis, and serum was positive for ketones. The story was that he had increasing thirst and urination for more than three months. In early June he had a tonsillectomy for enlarged tonsils and his symptoms worsened greatly afterwards, but no one had checked his glucose. He had been to see his otolaryngologist earlier that day because of a very sore throat, a follow-up examination to the surgery. He was now feeling very ill, and he was short of breath, very thirsty, and his wife was very alarmed. The busy doctor saw him, diagnosed oral thrush, and sent him away to come back the following day. However, his wife insisted he go to the Emergency Room, at which time he was admitted and diagnosed as having diabetic ketoacidosis and severe hyperglycemia. He improved after eight hours of intravenous insulin and fluids, and the therapy was continued throughout the next day.


The patient improved rapidly without complications, but an inspection of the hospital record revealed that he entered the hospital, had a surgical procedure, and left without even a single glucose determination being obtained. The patient also was seen in the doctor’s office and no notice was made of what had to be a very dehydrated tongue.


This case can be considered a "close call", or a "near miss", and nearly resulted in a catastrophe. But it illustrates that best practices were not followed here. The error was not to consider the problems of the patient with a broader focus. It also points to the need for some overall basic standards to be in place for inpatient care.


For example, I would make a case that a glucose level should be obtained on each hospitalization for any patient who has a family history of diabetes, is over the age of 65, or has other risk factors for diabetes or pre-diabetes. It would also be good for each specialty society to create guidelines or consensus papers regarding the constellation of symptoms or signs that are indicative of a condition not in their specialty area of expertise, which is not uncommon, where delay in diagnosis could lead to catastrophe. In a sense, these would be decision aids – practical points to encourage physicians to look for common but serious problems that are often encountered by their physicians, but not problems that they normally learn how to treat.


Coordination of care of patients does not mean that any one physician is solely responsible for all of the needs of a patient. It is more about teaching each of the physicians involved in the care of a patient with multiple problems enough about the other conditions that they do not treat, that they can serve as an alert to other physicians and thereby prevent the severe delay that put this patient at great risk. Our patients deserve no less.





Is there a link between Lantus (insulin glargine) and cancer? A commentary on reports from the FDA and the EASD on the possible association.


2009-07-02 15:42:54

By: Richard Hellman, MD, FACP, FACE

Recently, four articles and one letter were published in Diabetologica that set the diabetes world into a state of "high alert". The four articles, all observational studies and retrospective population studies, looked at the possible association between insulin glargine and cancer, and showed evidence of a possible association. The FDA has reviewed the data from the studies published in Diabetologica, and concluded that more data collection is needed.


The initial paper by Hemkens et al, was received by Diabetologica on 29 August 2008, but was not accepted until 26 May 2009, and was published only after the journal asked for and received papers from three other registries so a cross-continent, multinational comparison could be obtained. Although the study by Hemkens examined administrative data from more than 127,000 patients, the study was weakened by the fact that there was no randomization of the studied groups, the study was relatively short in duration, and, in this retrospective study, there was a limited ability to directly measure any allocation bias. In fact, the groups under comparison were known to be dissimilar with respect to both weight and age, and very possibly numerous other factors as well, factors that could well affect the relative incidence of cancer.


In the Hemkens study, they noted a dose-dependent relationship between the insulin glargine and the relative risk of cancer, but other observational studies published at the same time did not confirm this relationship. In fact, in the study from the Scottish Diabetes Research group, overall insulin glargine use was not associated with an increased risk of all cancers in Scotland over a four year time frame. But in one subgroup, patients only on one insulin, insulin glargine, there was an excess incidence of both breast cancer and overall cancer. However, the authors concluded that this finding was due to an allocation bias rather than an effect of insulin glargine itself.


Interestingly, in one of the studies reported, there was evidence that with the concurrent use of metformin, the increased cancer risk noted with insulin glargine was not present. This was interpreted as evidence of the cancer protective effect of metformin. In an accompanying editorial, there was discussion of the proposed mechanism by which metformin may reduce the risk of cancer.


Observational studies of the kind reported in Diabetologica are prone to great difficulties in interpretation because of the problems in dealing with confounding variables that are unmeasured. The results of the four studies, when compared, are conflicting in a variety of ways, leaving us in a bit of a quandary.


A much smaller study by Rosenstock and Fonseca et al, published as a letter in Diabetologica, gives data from a randomized trial that compared insulin glargine and NPH, and in a secondary evaluation, examined the incidence of cancer in the two groups over a four year period. They did not find any increased incidence of cancer in the group that was on insulin glargine.


On the other hand, there is clear evidence, from experimental laboratory studies, that insulin glargine has a six to eight-fold increased affinity for the IGF-1 receptor as compared to human insulin. This is associated with an increased mitogenic potency of the same degree (Kurtzhals et al.). Whether this increased affinity to the IGF-1 receptor translates in vivo to an increased risk for cancer was unclear before the articles in Diabetologica, and is still unclear.


What is clear is that high risk populations may deserve special scrutiny, since if in fact there is an increased risk for cancer due to insulin glargine above the well-known mitogenic potency of insulin in some experimental systems, the patients at highest risk for cancers may be the most at risk. Accordingly, there is a need to examine the data that has been obtained from patients on long-term use of insulin glargine and scrutinize it more carefully for any evidence of risk for malignancy.


It would, of course, be unethical to perform a randomized prospective study at this time to examine the questions raised by the studies on insulin glargine, but there may be other methods that will provide useful data that is more powerful than the aforementioned studies. Perhaps the relatively new technique of Mendelian Randomization may be utilized here in order to provide more cogent data. There may be yet other creative ways that are scientifically sound to provide new and useful information for us and for our patients.


It is clear that individualization of patient therapy will remain key. It is unlikely that anyone will have an appetite for the use of insulin glargine during pregnancy. But there may be other groups of patients in which this insulin analogue should also be avoided. At this time, however, the data are too scanty to be sure, and there is no convincing evidence that insulin glargine is unsafe as we are currently using it.


However, we hope that better data will be forthcoming soon so we can see whether Lantus (insulin glargine), which has had an exceptional safety record, will continue to be one of the mainstays of modern insulin therapy, or whether this new information regarding increased risk of cancer with the use of insulin glargine will be validated by more definitive studies. Stay tuned.





The July 2nd NEJM Correspondence regarding the NICE-SUGAR study – Were the better outcomes of the Van Den Berghe studies due to more accurate glucose meters used in point-of-care testing? Is patient safety for diabetes inpatients at risk in our hospitals with the methods in common use today?


2009-07-07 14:46:17

By: Richard Hellman, MD, FACP, FACE

A series of very provocative letters were published in the NEJM1 regarding the results of the NICE-SUGAR study, but the most interesting was the one by Greet Van Den Berghe. In her letter, she pointed out four possible explanations for the discrepancy between her published studies in the NEJM in 2001 and 2006 and the NICE-SUGAR study. She reminded readers that the differences in glycemic targets between the NICE-SUGAR study and the Leuven studies were considerable, in particular, the upper target in the NICE-SUGAR study was 140-180mg% (7.8-10.0mMol/L) vs. 180-215mg%(10.0-11.9mMol/L) in the Leuven studies, a significant difference. She also pointed out that the caloric intake was very different between the patients in both studies, with the NICE-SUGAR patients being on hypocaloric diets. She also pointed out that the early withdrawal of care (after a median of six days) was not balanced, and could have played a role.


But the strongest statement Van Den Berghe made was that she was surprised that a variety of glucometers were used in the NICE-SUGAR study. She stated that, with respect to the glucose meters, "…most of which were unsuitable for this purpose…" She felt that, as a result of the poor choice of glucose meters, "undetected hypoglycemia, large fluctuations in glucose levels, and possibly hypokalemia were tolerated or even induced." She goes on to say, "Such errors may have contributed to excess ‘cardiovascular' deaths, in the absence of differences in organ failure."


The response from the NICE-SUGAR authors, Finfer and Chittock, was most interesting and somewhat surprising. While they disputed the point by Van Den Berghe that the correct caloric levels in critical care were known, and felt the early withdrawal of 3.8%of the patients to palliative care was not significant, when it came to glucose testing methods, they said, "We do agree that more accurate systems for blood glucose measurement are required." This appears to be in full agreement with Van Den Berghe's statement on meters. But Finfer and Chittock went on to say that they thought it was unproven whether more accurate blood glucose measurement can make intensive glucose control safe and beneficial and suggested that it be studied in yet another trial.


I am not sure whether such a trial would pass an ethical review. Imagine for a moment whether you would be comfortable caring for a critically ill patient in which the serum potassium was accurate only to plus or minus 20%, the current FDA standard for glucose meters on the market today. In this scenario, you are faced with a patient with a bradyarrythmia and recorded potassium of 5.5meq/L, which you know, based on the accepted levels of precision and accuracy, could be from 4.4meq/L to 6.6meq/L. Would you find this level of accuracy and precision acceptable? In contrast, most hospital laboratories have an accuracy of about 2-4% at present, a more acceptable range. Is it fair to knowingly use glucose testing equipment with this level of accuracy in critical care when more accurate devices, such as used by the Leuven group, are available?


The comments by Van Den Berghe highlight the concerns that I have commented on in my editorial of February 25th, and most recently, on May 29th 2009,


I think it is time for us to call upon the FDA to work on improving glucose testing devices, but also for professional societies to take a stance on what are acceptable standards for accuracy and precision for patients who are critically ill and need point-of-care glucose testing.


There were other very interesting points made in the NEJM collection of letters on the NICE-SUGAR study. The comments of Pei and Yi, from China, point out what I think is another flaw in the NICE-SUGAR analysis, the long delay before there were statistically significant death rates, suggestive of post-randomization events playing a significant role in the observed outcomes. The study by Mohammedi, Roussel, and Marre, from France, point out that the disparity of frequency of corticosteroid use may have been a potential confounding variable, but the key point, I think, is the point by Van Den Berghe on the need to use the most accurate methods of glucose analysis when protocols are driven by the measured bedside glucose level. There is a need for all professional societies to take a stand on this important issue, even if it is unpopular among the manufacturers of these devices.





AACE and the FDA


2009-08-14 16:34:40

By: Richard Hellman, MD, FACP, FACE

AACE's link with the FDA began almost three years ago, when the AMA, led by Dr. Michael Maves, their Executive Vice President and CEO, convened a meeting between senior FDA staff and a number of representatives of the medical specialty societies and began a series of meetings on topics of great importance to medicine. During the past several years, in my role as the FDA liaison, the communication and the dialogue with the FDA continued both in this format and with direct communications with the FDA. During my period as President of AACE, from 2007-2008, there was a series of new initiatives that were considered. Some of these have been discussed in the past year in the editorial pages of the AACE Patient Safety Exchange. The FDA is a central figure in drug and device safety in the United States, and together with AHRQ, the Joint Commission, NQF, and the Physician Consortium, has figured prominently in the issues discussed on this website during the past year.


I am very gratified by the announcement yesterday of the FDA alert regarding glucose meters and interfering substances. They point out, however, that there have been 13 deaths because of inaccuracy due to interfering substances, six since they first notified physicians and hospitals about the risk. This loss of life because of the lack of focus by the medical community needs to stop. These deaths are all preventable. The issues of glucose meter accuracy and precision and the threat of interfering substances have been discussed at length in previous pages on the Patient Safety Exchange website, beginning on February 25th and most recently in my other editorial posted today. In that editorial, I delineated the key role AACE played in urging the FDA to re-consider their position regarding the low standards of accuracy and precision regarding the glucose meters and pointed out again the urgency of dealing with the not-generally-known risks of the GDH-PQQ enzyme monitoring technology that is used in many popular glucose meters.


The announcement yesterday by the FDA to specifically remind both the public and health care facilities of the risks of meters which use this technology is an important step forward for the agency. I am also pleased that the FDA is extending their reach not just to AACE, which they did in May 2009, but now, more recently, to other key groups, such as the American Diabetes Association and the Endocrine Society. I think this is indicative of the efforts they are making to connect and disseminate their findings.






Preliminary Report of the NICE-SUGAR study:

Hypoglycemia and what it may mean for patient safety


2009-02-23 18:22:16

By: Richard Hellman, MD, FACP, FACE

The optimal treatment for the hyperglycemic patient in an ICU setting has been hotly debated for many years. On one side are those who propose that normoglycemia is the goal of treatment of hyperglycemia, and on the other side those who propose that the case has not yet been proven – that the risk of trying to do so, particularly the higher risk of hypoglycemia, may outweigh the benefits. Perhaps the strongest evidence in favor of using insulin infusions and setting goals of normoglycemia is from Greet Van Den Berghe and her colleagues in Belgium, who performed a large randomized controlled trial to assess the optimal treatment of hyperglycemia in the SICU setting1. They found that maintenance of a goal of normoglycemia on patients in SICU settings for five or more days reduced mortality by 32% overall, 44% in those with sepsis, and significant reductions in both morbidity and hospital costs also occurred. A very recent study in Lancet by her team, Vlasselaers et al, published in The Lancet on February 14th, 2009, extends their evaluation by finding a significant benefit in reducing morbidity in the pediatric age group, but with a significant increase in the frequency of hypoglycemia2.


The hypoglycemia appears to be the main barrier to widespread adoption of intensive glycemic control in inpatient care. There are many questions still unanswered. Can we reduce the high incidence of hypoglycemia noted in so many published studies? Are we using the wrong tools? Are the instruments we use to measure glucose introducing errors which result in their suboptimal accuracy and precision? These questions and others will be the subject of this and subsequent editorials.


Two important studies have weighed in on this topic in the past weeks. The most important, but perhaps the most preliminary information is from the NICE-SUGAR study group . Their preliminary report of the first 100 hypoglycemic events can be found on the ANZCA website, at


The NICE-SUGAR study is a large multinational study that evaluated 1838 patients in ICU settings. They compared an intensive glycemic control strategy, which utilized continuous insulin infusions, with a strategy that was less intensive, with less stringent glycemic goals. Their findings were a surprise to many. The most common explanation for the adjudicated causes of the hypoglycemic events was clinician error, defined as failure to follow the computerized treatment algorithm, as for example, inappropriately infrequent blood glucose monitoring. Clinical errors alone explained 37% of the hypoglycemic events. The next most common cause was decreased nutritional intake, which explained 24% of the hypoglycemic events.


Put another way, their preliminary data indicates that although their algorithms to achieve normoglycemic were sound, if they were not implemented correctly, hypoglycemia commonly resulted. It is yet another example as to why it is not enough to have an excellent plan, but we need to make sure the implementation of this is correct. Without adequate training and monitoring of how the plan is delivered, suboptimal results often occur. This data provides a potential answer to the question as to why others have not duplicated van Den Berghe’s results. The failure to do so may be more a systems problem than a failure of concept.


Another study, published online on February 5, 2009 in Diabetes Care by Blaha and colleagues, may also be relevant. In this study, performed in Prague, the authors compared three protocols for tight glycemic control in cardiac surgery patients. They noted a relatively lower rate of hypoglycemic events overall in the three choices of algorithms, and made a point that the experience of the ICU staff with the protocols and the appropriate implementation period may have been one of the reasons why the quality and safety of glucose control in their study was significantly better than most of the previously published studies. However, by design, this study focused only on their chosen outcomes, glycemic control and freedom from hypoglycemia. It is therefore not comparable to the NICE-SUGAR study due to the lack of morbidity and mortality data.


Nevertheless, both the data from Blaha study and the preliminary data from the NICE-SUGAR study strongly suggest that failure to duplicate the positive findings of the Van Den Berghe studies may often be due to a failure to pay adequate attention to the need to adequately train and monitor the work of the personnel who are managing the insulin infusions. Creating adequate systems of care for this complicated therapy is a must, and results may be much better once there is more consistent attention to this important aspect of care.



    1. Van den Berghe G, et al. Intensive Insulin Therapy in Critically Ill Patients, NEJM 345: 1359-1367
    2. Vlasselaers D, et al. Intensive insulin therapy for patients in paediatric intensive care: a prospective, randomized controlled study, Lancet 373:547-556




Editorial on FDA alert on confusion between propylthiouracil (PTU) and purinethol (mercaptopurine)


2009-01-09 13:41:28

By: Dr. Richard Hellman

The FDA has just released an important alert from the Institute for Safe Medication Practices (ISMP) of yet another example of errors in drug administration due to confusion between two very different agents, a drug used to treat hyperthyroidism, PTU, and an agent used in oncology, purinethol (mercaptopurine).


The report delineates not just how the errors occurred, with a description of the harm the errors caused, but also concrete steps that we can take to reduce the likelihood of such an occurrence happening to our patients. Once again, we see the hazards of hand-written prescriptions and the value of electronic health records and order writing. But the ISMP also makes clear the importance of engaging the patient in identifying potential examples of drug confusion, and the need to avoid abbreviations because of the ambiguity they create. Double-checking of medications has been an important safety standard for a long time, and it is particularly important when dealing with easily confused medications.


The careful reader may note that I used the term of "yet another example" in the first paragraph. The reason is that when we instituted an electronic medical record with electronic prescribing in our practice in 1999, a patient with diabetes and breast cancer shared a story with me. She had gone to her oncologist after the surgery and he had prescribed an oral anticancer drug for her named cytoxan. After taking her new anticancer drug for two months, she felt increasingly ill, shaky, heat intolerant, with palpitations and profound muscle weakness. When checked, she had evidence of a suppressed TSH and slightly lower FT4 levels, but elevated T3 levels.


The mystery was solved when her prescription was examined. The label did not read cytoxan, but instead was cytomel, or triiodothyronine (T3). The pharmacist, unable to read the oncologist's prescription correctly, filled it as cytomel 50mcg daily, and she developed iatrogenic hyperthyroidism, with a two-month interruption in her anticancer regimen. With the cessation of the cytomel her symptoms disappeared, but the patient now understood clearly how medical errors can play a pivotal role in her health. She stated that it is very reassuring to her to see a system of care where handwriting confusion can be largely eliminated. She is now a great fan of electronic records, which she sees as a protector of her safety.


Confusion between drugs is inevitable because most physicians do not have an encyclopedic knowledge of both generic and brand names of medications, and the volume of new agents is rapidly increasing. There is a sound scientific basis to advocate the use of electronic aids and working collaboratively with pharmacies to reduce errors in drug administration. I hope that this FDA alert will remind us again how easily errors can develop and the value of using sound practices to promote patient safety.


Preventing Medical Errors - More Mix-ups between Propylthiouracil and Purinethol






Joint Commission Safety Alert regarding Health Information Technology


2008-12-23 16:05:42

By: Dr. Richard Hellman

On December 11, 2008, the Joint Commission released a sentinel event alert regarding implementation of health information technology (HIT) and converging technologies – the interrelationship between medical devices and information technology.


This very thoughtful document highlights the many ways that poorly designed or implemented HIT products can lead to medical errors and serious problems in patient safety. Although there is much less data on the incidence of adverse events directly caused by HIT, they report data from the United States Pharmacopeia’s MEDMARX database which includes medication error reports, and approximately 25% of the 176,409 medication error records for 2006 involved some aspect of computer technology as at least one cause of the error. Examples included mislabeled barcodes, poorly-designed information management systems and unclear or confusing computer screen displays, but there were many others including those related to computerized physician order entry systems (CPOE) as well.


Unfortunately, often HIT interventions are not done with the inclusion of the front-line clinicians in either the planning or the initial implementation testing, and the results are often disastrous. The Joint Commission report delineates many ways in which the failure to include the front-line clinicians results in errors, which in retrospect, should have been perfectly predictable, given the importance of those at the point of care and the complexity of the work they do.


The Joint Commission has suggested 13 actions to be considered prior to implementing an HIT solution in the health care workplace. These suggestions include the examination of workflow processes prior to any implementation, and improving those processes first instead of relying upon the belief that the HIT solution will “cure all ills”. They stressed the need to test and check what is being proposed, and to take a hard-nosed look as to whether the implementation has improved the processes or merely created new problems, each of which need new solutions.


In summary, the Joint Commission is to be commended for looking beyond the “hype” of information technology solutions and embracing HIT for what it can do to improve health care safety, while recognizing the new problems in patients safety HIT may cause. The Joint Commission suggests general strategies for how to best design and maintain HIT systems to maximize their value and minimize the problems they may bring. HIT is not a panacea for problems in patient safety, it is merely a tool that, if well designed and implemented, can improve patient safety, but at the same time, even the best HIT products present novel challenges and complexities that must not be underestimated if we are to maintain a safe environment for our patients.




Editorial regarding interview with Carolyn Clancy, MD


2008-12-08 14:41:11

By: Dr. Richard Hellman

The Patient Safety Exchange is proud to begin what we hope is a long-running series of interviews with experts in the field of patient safety. It is fitting that the first interview is with the Director of the Agency for Healthcare Research and Quality (AHRQ), Dr Carolyn Clancy.


Dr Clancy is a general internist and health services researcher by training, and has worked at AHRQ for many years. She was named the acting director in 2002, and has been the Director of AHRQ since 2003. Under her leadership, the agency has made major contributions towards the development of an evidentiary base for the field of patient safety research and has been a center for some of the soundest and most innovative approaches to improving patient safety.


She is widely published, and has received numerous honors, including being elected to Master of the American College of Physicians in 2004. She is also a member of the Institute of Medicine.


In her answers to the five questions I posed, she focused on information meant to be immediately useful to the practicing physician, and provided links to the AHRQ resources that would be most helpful. I note that in her response to the fifth question, the question that asked about the importance of a coordinated approach to the care of the diabetic patient, she made the point that a survey of the AHRQ Medical Expenditure Panel found that only 9.8% of those with Type II diabetes have only diabetes, the rest having one or more additional chronic illnesses. She pointed out the high frequency of errors that occur during transitions between settings of care or between providers, and also emphasized the importance of patient-centered care. She then emphasized the need for team members, endocrinologist, primary care physician, pharmacist and others, to work together with the patient, their family, and the rest of their support system.


I could not agree more with her comments. Many experts, including the founding Director of AHRQ, the late Dr John Eisenberg, have pointed out the high level of danger that occurs when providers do not communicate well with each other. Key information and medical orders may be crucial to the safety of the patient, and when one or more providers are not aware of the critical information, the patient is put at great risk.


The dilemma for endocrinologists will be what their role should be in the patient-centered diabetes care of the future. Clearly, their role should be allowed to vary, dependent not only on the interests and skills of the endocrinologist, but also many other factors, including the wishes of both the patient and the other physicians involved in their care.


Some endocrinologists may wish to be the coordinating physician for some or even most of their diabetic patients. Other endocrinologists much prefer never being the coordinating physician, but to be always in a consultant role. Regardless of the choice of the endocrinologist, it is clear that a better understanding of how errors in care occur, and how to develop a culture of safety in their practice, will prove to be most helpful for the patient. In this all too brief interview, Dr Clancy provides us with useful tips and pointers, and the rationale for the changes in work-flow in our practices that will most promote patient safety. We thank her for so generously making herself available for our first formal interview.




The Obama Victory - How it May Affect Patient Safety Efforts in the US


2008-11-07 10:43:34

By: Richard Hellman, MD, FACP, FACE

Even those who did not vote for Senator Obama understand that his victory was of historic importance in the US, not only in repairing racial relations, but in choosing a direction of governance at a crucial time in our history. Facing two wars, the threat of global terrorism, a rapidly unraveling health care system, staggering debt, the world climate at risk, and the most serious fiscal crisis since the great depression, it is no understatement that many considered this Presidential election as the most important in a very long time.


The outpouring of joy and optimism regarding the election of Barack Obama has been stunning. I had not seen such a reaction in our country since the election of Ronald Reagan in 1980. When he said, "It's morning again in America", people wept. On this past election night, when Obama said, "Yes, we can", people wept, too.


But many still view the incoming administration with trepidation. There is a need to ask whether amidst the chaos, the well-recognized problems of gaps in quality and safety in our increasingly dysfunctional health care system will receive much attention by the incoming Obama administration. The problem is not whether people support improving quality and safety. It is like mom and apple pie - we are all for it. Instead the issue is where scarce resources go, and whether people understand that if we neglect quality and safety issues, no matter how much money we spend on access and new technologies, we still will have excessive morbidity, mortality, and costs, for patient safety is not an intrinsic property of complex systems. In fact, more complex systems merely are higher at risk of novel errors and injuries.


Patient safety efforts are well under way in the U.S. and the global medical community is very interested in how we are approaching the task. Led by the recent efforts of the Agency for Healthcare Research and Quality (AHRQ) in the U.S., and efforts by the Food and Drug Administration (FDA) as well, there is a concerted effort to make quality and safety a higher priority than it has been in the past. But there is a price tag to providing better training, education, and Information Technology (IT) infrastructure. Double checks, detailed self-analysis, and providing quality data for scrutiny makes sense, but costs both time and money. Who will foot the bill? In past years, the answer was clear - not the payors. Whether government or private sector, whether Medicare, Blue Cross, or a for-profit managed care company, "the game" was to shift costs to someone else, preferably the provider, especially the physician. This cynical approach has driven many physicians out of direct primary care and from disease management. It has made quality and safety improvements nearly impossible. Will it happen in an Obama administration?


So far, there are grounds for cautious optimism. Having had the opportunity to speak with the senior health advisors for both Senator Obama and Senator McCain, the sense I have is that Senator Obama’s team is very aware of the gaps in care in public health, and the multiple failures in quality and safety areas in health care in the U.S. Their initial plan, although painted only with broad strokes and insufficient in detail, appears to emphasize disease management, coordination of care, and providing financial incentives regarding health IT. I think they are most supportive of focusing on programs that work, and do not appear to be particularly ideological when it comes to the type of health care entity needed to get the job done. They appear to recognize that high quality organizations may look very different in structure, and yet accomplish the same result for the patient.


But will the money be there? Will the allocation of resources be up to the task, and will each of the entities do their part? I hope physicians and hospitals realize that the days of saying, "Give us the money, trust us to do our part, and then it will work out fine" are gone. Neither payors nor health policy leaders find that point of view as very credible, for there are too many bad things that happen at the best institutions, even with the most sophisticated health care systems. Complexity is hardly a guarantee of success and patient safety. On the contrary, complexity may increase the probability of novel failures.


On the other hand, I think there is an increasing awareness by government planners that they do not have the skills to deliver or direct the high quality innovations that will increase patient safety efforts. Instead, there is an emerging awareness that a better plan may be a mutually cooperative effort where the payors provide the resources, both that are used in transparent, accountable, health care systems with sufficient flexibility to innovate improvements in quality, and to provide a continuous training environment for the patient-centered health care team. Continuing dialogue with the Obama health care team may result in a more practical approach to improving quality and safety in health care. We need to avoid the trap of thinking that a good plan of care is equivalent to delivering that plan flawlessly, in a timely fashion, and achieving the patient outcomes that our public deserves. Effective patient safety initiatives should make few assumptions, but instead rely upon data to show us whether we are hitting the mark, helping our patients with our efforts, and doing so in a fashion that avoids inadvertent harm. I hope the incoming administration continues to engage the quality and safety community and focus on improved and safer patient care. These discussions can only benefit patient care and safety efforts.







The JDRF Study on Continuous Glucose Monitoring


2008-09-08 14:23:42

By: Richard Hellman, MD, FACP, FACE

Today, at the European Association for the Study of Diabetes, the researchers for the JDRF study group presented their findings, just published on-line in the New England Journal of Medicine. The trial is almost certainly to become a landmark study, a carefully done randomly controlled trial (RCT) by the JDRF study group on the usefulness of continuous glucose sensing in a carefully selected population of patients with Type 1 diabetes, all of whom were either on external insulin pumps (CSII) or on at least three insulin injections daily (MDI). The primary end-point was to achieve a reduction of HbA1c of at least 0.5% over the six month period and one subgroup, the patients age >25 years showed this result, reducing their mean A1c levels by -0.53% with a p-value of less than 0.001, which is highly significant. Even more impressive was the fact that there was no significant increase in hypoglycemia in this group despite the improvement in an A1c from a baseline of 7.6%.


The carefully constructed study was actually three studies in one. They studied and compared the results of the intervention of the CGM tool together with educational support and follow-up in patients with type 1 diabetes over the age of 25, and two other subgroups, those ages 8-14 years, who did less well, and those ages 15-24, who did least well. Another key finding is that the primary result correlated very closely with age, with a p-value of 0.003, and that the willingness of these patients to continue using the glucose sensor frequently varied with the group, with the most frequent use in the patients at least 25 years or older, and the least with those ages 15-24 years.


The article, which is very rich in detail, will undoubtedly result in many more articles to extend the information presented in the key article, published online today in the NEJM, but the research group clearly has much more analysis yet to come. For instance, they have not yet published any of the psychological data regarding the adolescents. We need to understand the barriers a provider will face in trying to improve on the A1c results found in those ages 15-24.


We do have some clues at this point. The correlation between how well even the age 15-24 year old group did with respect to glycemic control when they used CGM at least 6 days weekly was striking. The same was noted for the 8-14 age group. The subsequent analysis, not in the paper, showed that the group aged 13-17 had the most difficult time overall in reaching the A1c and adherence goals.


It was interesting that the data did not show a difference between those patients who used MDI therapy and those who used CSII, suggesting that the most important difference was the ability of well-prepared patients to take the real-time feedback on their glucose levels and adjust their behaviors and therapy to improve their glycemic control. These patients lowered their A1c and their risk of complications without increasing the incidence and severity of hypoglycemia. In fact, there were favorable trends in the lowering of the incidence of hypoglycemia.


It is also important to note that this study did not explore the usefulness of this technology with patients with type 2 diabetes, or those who are pregnant, or those who are aged. Also, patients with multiple severe co-morbid conditions were not included. Although many of us believe strongly that CGM has a very broad level of usefulness, this JDRF study did not provide data to support the broader use.


However, the most serious problem with the article, which I am told was a result of the length requirements of original articles imposed by the NEJM, was the relative understatement of what was required in time and resources to provide the care and education that was so successful in this trial. At this time, there is a serious disincentive for physicians and other providers to offer this intensive form of education, improved technological support, and follow-up. There is, at present, no payment offered for any of the time that one spends providing what the patient needs to know, and the follow-up that must be customized to the needs of the patient. The study investigators appeared to agree with this point, but it was not explicitly stated in the article. I hope this will be corrected in follow-up publications, because the dissemination of this advance in therapeutic ability, the use of CGM, which can be used to lower the A1c to ideal levels without increasing hypoglycemic rates, needs not to be hindered because of lack of infrastructure support for those providing the education and care, which should ideally be integrated for best results.


It is important for endocrinologists to help take part in the task of training not only our own members in how best to use this new technology well, but also to help train others. Endocrinologists, primary care physicians, nurses, dieticians, psychologists, and others of our community who are skilled in diabetes care and education, need to be mutually cooperative. We will need to work with key patient advocacy groups, to make sure the community of care for those with diabetes has the resources to make it economically feasible for us to help bring this advance to our patients and use it to their best advantage.





Additional Thoughts on ACCORD and ADVANCE Trials


2008-09-08 13:42:29

By: Richard Hellman, MD, FACP, FACE

The ACCORD, ADVANCE, and VADT trials are landmark studies that will continue to have a major impact for years to come, but each of the studies managed to reach conclusions without any explicit discussion of the role that human error plays in affecting clinically important clinical outcomes. People make mistakes. Mistakes with powerful therapy can cause injury. Patients with diabetes often have very complex treatment regimens, and those in the ACCORD trial had extremely complex therapeutic regimens. 52% of the intensive group had three different classes of medication, as well as an insulin regimen. Do we know whether the patients’ cognitive and emotional equipment was up to the task? Did they do their jobs well, or with errors, and if with errors, how often, and what were the consequences?


In no way does this reflect on the skill of the providers, who include some of the very finest clinicians and researchers, but the questions are unanswered and relevant, because in the real world, mistakes happen often.


A study published in May 2008 in Pediatrics highlights the frequency of injury or death of children and adolescents with inadequate education concerning their insulin pump. But the most important aspect is that much attention is paid to what constitutes the curriculum, but far less attention in the literature is being paid to "what is the best way to help make the patient be knowledgeable and functional enough to provide their own self-care safely and effectively?" Moreover, there is very little literature regarding methods of identifying when a patient cannot be counted upon to give the safe self-care needed, and how best to help coordinate the resources needed.


The bottom line from each of the trials seems to be that intensive therapy is more effective the earlier it is begun. This is not a new observation, certainly the DCCT data showed the same, but it is reassuring nevertheless.


The ACCORD trial certainly raises the question whether intensive glycemic therapy is safe for all patients. But it is not correct to extrapolate from the mortality data from ACCORD to all patients, and yet, despite the VADT and ADVANCE trials not confirming the mortality data, the ACCORD findings are very important.


I think the ACCORD data and conclusions can be rephrased as follows:

Intensive glycemic control strategies can increase undesirable outcomes. Although we do not know as much as we would like about the cognitive and psychological data from the ACCORD-MIND study, we will probably find that individualizing intensive glycemic control is both safer and more likely to be effective.


We need to study the frequency of human error in the care of diabetes, not only in providers, which is still far too common, but in the patient population, which is implicitly acknowledged to be significant, but is almost certainly a significant factor in adverse outcomes, as data from studies on hypoglycemia, death from DKA, and foot care have shown.


I look forward to seeing these subjects explicitly evaluated in subsequent trials.





Patient Safety in Diabetes Care and Intensive Glycemic Therapy


2008-09-04 10:38:20

By: Richard Hellman, MD, FACP, FACE

In June 2008, the results of both the ADVANCE and ACCORD studies became public. The results were published in the New England Journal of Medicine (NEJM) on June 12, 2008, presented at the American Diabetes Association (ADA) meeting, and widely discussed in the public media as well. The two trials, ADVANCE and ACCORD, are clearly landmark trials, both because of the high quality of the studies, and the importance of the results. Yet, the two show what appears to be conflicting conclusions.


The ADVANCE trial, a multi-center RCT study with 215 centers in 20 countries, involving 11140 patients with Type 2 Diabetes, compared the results of intensive glycemic management between an experimental group of 5,571 patients receiving therapy with a goal of an A1C of 6.5%, and achieved that goal with a median A1c of 6.5%. The authors concluded that the goal was achievable, safe, and beneficial.


The primary outcome of combined macrovascular and microvascular events was 18.1% in the experimental group, and the standard therapy group had a major outcome of 20.0% over the five year average study period.


The major benefit in reduction of complications was a reduction of nephropathy, 4.1% intensive vs. standard 5.2%, with a hazard rate of 0.79 and a P = 0.006. In contrast, there was a nonsignificant reduction in major macrovascular events with a hazard ratio of 0.95, p = 032. They did also note that severe hypoglycemia was more common in the intensive - control group (2.7% to 1.5%, hazard ratio 1.86, p = 0.001)


But the results of the ACCORD trial were strikingly different. As is now widely known, the safety committee of ACCORD concluded that the excess of deaths noted in the intensive therapy group was sufficiently alarming to require the cessation of the study for safety reasons. 257 patients in the intensive group died, compared to 203 in the standard group. The hazard ratio was 1.22, p = 0.004. They noted, however, that the primary outcome of macrovascular event was less frequent in the intensive group, 352 vs. 371 in the standard therapy group (hazard ration 0.9, p = 0.16). Also noted was a lower number of non-fatal myocardial infarctions in the intensive group.


Moreover, subgroup analysis suggested a clear benefit for intensive therapy among patients with an A1c <8.0% at baseline, and with patients with no previously documented cardiovascular disease.


There has been, so far, far too little attention paid to the fact that we do not yet have data from a very important sub-study embedded within the ACCORD, the ACCORD-MIND study, which is a careful and detailed analysis of the cognitive and emotional changes noted in the study patients, both during the 3.5 years of the intensive glycemic phase and continuing until 2010.


In contrast, the ADVANCE Trial did report that there was a nonsignificant risk reduction of cognitive decline of 2% in the intensive therapy group, but the most detailed analysis and data is likely to be found in the results of the ACCORD-MIND study, and that is not yet available.


But why should the ACCORD-MIND data be so important? Because most of the therapy in both the ACCORD and ADVANCE trials is prescribed by the research team, but actually administered by the patients themselves.


Until we know that the patient was as capable as they should be of doing what is asked of them accurately and in a timely fashion, we cannot exclude the possibility that the discordance between the macrovascular outcomes of two studies, ACCORD and ADVANCE, reflects a less-than-perfect way good strategies were either prescribed or administered.


To be certain, the patients in both arms of both studies, overall had lower death rates than what would be expected from previous epidemiologic data. The patients were certainly not placed in harm’s way. The ACCORD investigators were correct to stop the intensive therapy arm of the study, but we have insufficient data to be sure why the results were what they were in ACCORD. It is still plausible that the strategy used for all may have worked for the right patient. Moreover, the strategy of intensive glycemic control may still work if the method of implementation is optimized by adapting more closely to the emotional and cognitive characteristics of the patient.


We look forward to further studies to answer the questions raised by the findings of the ADVANCE, ACCORD, and VADT trials.





Depression and Diabetes and Patient Safety


2008-09-02 16:42:38

By: Richard Hellman, MD, FACP, FACE

My editorials usually begin at night, after work, when I sit down, take out a yellow note pad, and begin to scribble, in nearly undecipherable script, the outline of an editorial on a current issue in patient safety in the field of metabolism and endocrinology. Today is different. I have just left an examination room where my patient, a 72-year-old woman with diabetes, tearfully admitted that her sense of hopelessness and anguish, centered on the loss of her husband, kept her from doing what her distraught daughter begged her to do, which was to call us to be seen for the foot ulceration that began three weeks ago and is now complicated with a draining wound on her great toe with secondary cellulitis. The three-week delay will greatly increase the risk of a poor outcome in this lady with a prior history of severe peripheral neuropathy and foot deformity.


Depression affects outcomes in diabetes care in a variety of ways, some very obvious, such as this woman’s lack of ability to make the call for help in a timely fashion, and others more subtle, such as non-adherence to medication or lifestyle behaviors. There is now considerable data showing that depression significantly increases both morbidity and mortality in those with diabetes. It is also true that the incidence of depression in patients with diabetes is at least twice as high as in the general population. Moreover, depression in diabetic patients tends to be more often severe, recurrent, difficult to treat, and strongly correlated with complications of diabetes. Poor diabetes control is closely correlated with decrease in cognitive function, and this, in turn, greatly increases the risk of patient error or misjudgment, which can lead to morbidity and mortality.


This year, the PROSPECT trial was reported in Diabetes Care by Bogner, H et al.1 The trial showed the positive effect of a psychiatric intervention on reducing mortality rates in diabetic patients who were depressed as compared to a group of patients whose physicians knew they were depressed and were free to suggest whatever care they thought appropriate. The results suggest that a structured, more intensive approach to the treatment of depression saved lives.


Unfortunately, we probably do not always know in time whom to treat. Most physicians do not routinely screen their diabetic patients for either depression or cognitive dysfunction despite the fact that well-validated, easy-to-use tools are available to screen for both conditions. It is often only after an event that the depression, with or without cognitive dysfunction, is discovered.


Fortunately, the increased interest in the interaction of diabetes and depression is resulting in world-wide interest in dealing with these issues. An international group, led by Prof. N. Sartorius and representing leadership from many organizations, will be having a major conference in Europe early next year. Their effort will be focused on bringing together the fine work of many scientists, such as Dr. F. Snoek, Dr. A. Jacobson, Dr. P. Lustman, and many others who have already contributed to our understanding of the interaction between depression and diabetes.


It is clear that depression is a co-morbid condition that must be diagnosed and treated well if those with diabetes are to be able to have the quality of life they and their families deserve. We can and must do better so we can avoid the added morbidity and mortality risk that depression places on our patients.




1. Bogner H, Knashawn H, Post E, et al. “Diabetes, Depression, and Death”. Diabetes Care, Vol 30:12:3005-3010, December 2007





Editorial on Byetta (exenatide) and Patient Safety


2008-08-25 16:22:21

By: Richard Hellman, MD, FACP, FACE

The recent FDA alert regarding Byetta1 (exenatide) should be noted with concern. The mortality rates of severe acute pancreatitis in published reports is variable, but is usually reported to be over 10%. In the six cases the FDA reported of acute hemorrhagic pancreatitis, the mortality rate was 33%. In the previous FDA report on exenatide in 2007, none of the cases of pancreatitis led to death, although there were serious complications in several cases.


What we do not know – as is so frequently the case – is what the true incidence is of pancreatitis in patients who are placed on exenatide. Is pancreatitis on exenatide a rare phenomenon, or is it far more common than presently reported? We simply do not know the answer. There are many patients who are currently doing very well on exenatide, probably hundreds of thousands of patients at last count. However, exenatide often is not tolerated because of nausea and vomiting. We do not have data from post-marketing studies as to how often those symptoms, which may have preceded the cessation of the drug, were associated with a very mild episode of pancreatitis.


The New England Journal of Medicine's series of letters, "Exenatide and Rare Adverse Events"2, published May 1, 2008, is also relevant, since Dr. Pablo Cure, one of the authors, reported on a patient on exenatide who developed a gastric bezoar, a very serious cause of nausea and vomiting, another complication that would also necessitate cessation of exenatide. Although the NEJM response from Amylin, the manufacturer of exenatide, indicated that they knew of only one other case, I have another. A 51-year-old woman with type 2 diabetes of 12 years duration, without significant peripheral neuropathy and no prior history of gastroparesis, developed intolerance to exenatide after starting the medication in July 2005. She stopped the medication three months later, and two months after that an endoscopy revealed a gastric bezoar. I, as many of my colleagues, realized that the clinical information was important but did not take the time to report the case either as a case study in the literature or as a report to the FDA.


It is hazardous for anyone to extrapolate from the one case I report and estimate the incidence of gastric bezoars or severe gastroparesis that have developed as a result of exenatide therapy, but we clearly need more data. We also need more physicians to ask the question as to why their patient is not tolerating exenatide, obtain relevant data at the time of the symptomatology, and share the data that may indicate more serious complications such as pancreatitis and gastric bezoars, either by reporting the data to the FDA, or in the literature, or both. We need to know the frequency or rarity of these serious complications. We also need data regarding the comorbid conditions and demographic data on the patients who are having the more severe problems. It may be that exenatide is safe for nearly all patients but not for some patients because of their risk factors for complications, risk factors that may still be unknown.


I hope that the Sentinel system of drug reporting will increase the responses to the FDA, but we also have a responsibility, both as individuals and as a professional society, to share what have learned. The tools we use for our patients are often highly potent therapeutic agents, all of which have potential limitations on their optimal and safe use. Exenatide, a truly remarkable medication, whose mechanism of action is complex and yet incompletely understood, will need closer scrutiny and more complete post-marketing data. Once again, in hindsight, it is clear that a more extensive, detailed and expensive outcome study would have been most helpful.


It is likely that some hard negotiation between the pharmaceutical industry and our government will be needed, so we can do a better job protecting the safety of our patients without stifling drug development. Perhaps a private/public partnership to pay for the needed studies is the answer. Perhaps changes in the length of patents may be another solution, but it is clear that our present system of drug approval sometimes not only puts our patients at risk, but also results in events that unfairly label drugs that are useful as "bad", when the fact is that they are good drugs for many, but not all. A more robust front-end evaluation of drug safety may be in the interest of all.








Insulin Pens and Safety


2008-08-04 16:19:36

By: Richard Hellman, MD, FACP, FACE

The Institute for Safe Medication Practices (ISMP) released an article recently that highlighted potential safety problems that are being seen in hospitals when they switch the delivery of insulin subcutaneously from insulin that is drawn from multiple-dose vials to insulin pens. The details of the ISMP Medication Safety Alert were included in an FDA Patient Safety News Alert, #78, which is attached. The study is noteworthy on its own, but the implications for all of us are much broader.


The ISMP properly points out that although insulin pens have many potential advantages, including the potential for achieving both greater accuracy and precision of the insulin dose given, there are potential hazards as well.


For starters, needle stick injuries occur because the pen makes it difficult to see the injection site, and some needles do not have needle guards. Some nurses also will deliberately remove the needle guard because they find it cumbersome.


Moreover, the nurse who is not careful may forget that it is not safe to reuse the pen for multiple patients. Biological materials from the first patient may appear in the insulin pen, and then be injected into the next patient. And at time some nurses have held the pen upside down and misread the dose of insulin, giving an incorrect dose.


To summarize, it is not safe for a hospital to switch from insulin vials to insulin pens without putting their nurses through a rigorous program of education, because in real life, many unexpected errors have occurred. It is not enough to merely point out how these devices work and how they are used, it is at least equally important to recognize how humans, even experienced health professionals, can use them improperly and the harm that can result.


But this editorial is not just about making sure that all of the hospital nursing staffs are using these devices properly, but about much broader issues. Are your patients getting adequate education when they are put on an insulin pen? Did they learn what they needed to know? And even more importantly, what are they doing now when they use an insulin pen?


If you do not know the answer, it would be wise to check how they are doing. You may be pleasantly surprised, but on the other hand, you may be appalled by what you find out.


This is just one example of why patient safety issues are so important. We often take for granted far too much about how people, both providers and patients, are doing. For most of us, especially when we are distracted, do not learn flawlessly, nor remember forever. Yet for us to help our patients as much as they need us to, we must cover the last point, the "slip twixt the cup and the lip", in order to make sure that our patients are getting the educational intake that they need.


Additional Information

FDA Patient Safety News - Potential Problems with Insulin Pens in Hospitals





The Veterans Administration Diabetes Trial (VADT) and Lessons for Patient Safety


2008-07-14 14:47:06

By: Richard Hellman, MD, FACP, FACE

Three important trials were presented last month at the American Diabetes Association’s 68th Scientific Sessions. Two of the trials, the ACCORD and ADVANCE, were published in the June 12, 2008 edition of the New England Journal of Medicine, but at the time of the presentation, the authors of the Veterans Administration Diabetes Trial (VADT) had just closed their database and had not yet completed all of their analysis of the data. But even with the preliminary data, some important conclusions can be drawn.


The first conclusion of the VADT was an important negative result. The researchers could not show that, in the setting of improved control of cardiovascular risk factors such as blood pressure, lipid control, and lifestyle training, intensive glycemic control reduced the incidence of major cardiovascular events.


There were, however, some intriguing findings. For example, a significant reduction in risk of cardiovascular events was found in the patients with a shorter duration of diabetes. After 15 years’ duration of diabetes, the risk associated with intensive therapy, as measured by a hazard ratio, was >1 and tended to increase with increasing duration of diabetes.


Another unexpected finding was regarding hypoglycemia. Overall the hazard ratio for having a major cardiovascular event in patients who had a major hypoglycemic episode in the previous three months was 2.062, but in those on standard therapy the hazard ratio was 5.855 vs. only 1.277 in those on intensive therapy.


In contrast, it was hardly surprising that coronary artery calcification (CAC) proved to be very strongly associated with a major coronary event, an association even stronger than a prior history of cardiovascular event. For those with a CAC score >400, the relative risk of a cardiovascular event was 7.5 times that of the entire study population. Only 6% of all patients studied had a CAC score >400. Certain complications were associated with a disproportionally high likelihood of having a CAC score >400. For instance, 80% of patients with proliferative diabetic retinopathy had a CAC score >400.


The authors evaluated the relative cardiovascular risk associated with rosiglitazone use. This was not an easy task, because although the majority of patients received this agent, there were many differences in timing, duration, and dose of rosiglitazone therapy. Consequently, the biostatisticians chose a case-control study method of analysis. They reported that their analysis showed no increased cardiovascular risk for rosiglitazone, but admitted that this analysis might be less compelling because the VADT was not originally designed to study rosiglitazone, and as a result, the statistical technique best suited for the analysis, the case control study, did not yeld as high a level of evidence as a randomized controlled trial would produce.


But do the VADT results provide guidance in our efforts to improve patient safety in diabetes therapy? The answer, I think, is a qualified “yes”. The implication of the first analysis of the VADT is to make clearer that the strategy used in therapy must be tailored to fit the needs of the individual patient.


As an example: before beginning intensive therapy on 65-year-old diabetic patient of 25 years duration of diabetes with proliferative diabetic retinopathy and prior recent history of severe hypoglycemia, a thorough assessment of the patient’s true risk and status regarding cardiac risks is in order. Does this patient have high grade coronary artery disease? Does this patient have the ability to clearly see what they are taking? Can they sense hypoglycemia, or do they have hypoglycemic unresponsiveness? Does the patient have associated mild dementia or depression? What about the support from family and companions, is that sufficient regardless of the strategy offered to the patient?


The data on hypoglycemia has surprising implications. Often, physicians may assume that if intensive therapy is followed by an episode of hypoglycemia, then standard therapy might be a safer choice. The VADT data, however, showed that the risk for a major cardiovascular event after a major hypoglycemic event was much higher in the standard therapy group. The reason why is unclear. Could there be, for instance, a subset of those on standard therapy for which standard therapy may be more dangerous? It may be that some patients with previous episodes of severe hypoglycemia may benefit from either a more intensive approach or a different monitoring strategy. More analysis of these preliminary findings is urgently needed.


The VADT will, along with the ACCORD and ADVANCE trials, be recognized as landmark studies, but from the point of view of patient safety issues, these studies may well mark the entrance of major research efforts into just how best to implement our more fundamental research insights.


Clearly, strategies for glycemic control need to be individualized. The challenge for the clinical investigators will be to delineate enough of the important barriers to and risks of successful implementation of core strategies so that when we design long-term outcome research studies, we will be able also to analyze the relative importance of factors crucial to successful implementation of the therapeutic plans. To the patient, both attention to the barriers to and risks of implementation of a core strategy, such as intensive glycemic control, may make all the difference between success and failure. Our patients must be able to comprehend and consistently apply what they have been taught in order to get the positive clinical outcomes they need. And we need to provide what they need in education, training, and guidance in order to help them succeed.




A First Step Toward Improving Drug Surveillance – The Sentinel Initiative


2008-05-24 15:47:15

By: Richard Hellman, MD, FACP, FACE

The Food & Drug Administration (FDA), in conjunction with the Department of Health & Human Service (HHS) and the Centers for Medicare & Medicaid Services (CMS), has just launched an effort, the Sentinel Initiative, to develop a post-market drug and device surveillance system that relies on electronic databases. In order to underscore the importance of this venture, the leadership of HHS, CMS, and FDA announced their plans at a joint press conference on Thursday, May 22, 2008. This followed the announcement of a CMS rule giving CMS (and the FDA) the authority to look at quality and safety concerns by querying the Part D Medicare prescription drug administrative database.


The beginning of the Sentinel Initiative is an important step, but alone, will not solve the problem regarding our woefully inadequate post-approval drug surveillance. A successful effort will take years of work and much more in the way of resources. But what the Sentinel Initiative does is potentially very important and we need to understand it.


The status quo has been that post-approval drug surveillance at the level of the FDA depended upon adverse event reporting that came from multiple sources, from device makers, pharmaceutical companies, and by physicians, nursing homes, and hospitals, to name just a few. The problem has always been the incompleteness of the data sent to the FDA. All too often, the report sent would have neither the correct diagnoses, nor the co-morbid conditions, nor other important and highly relevant clinical data. But this was better than the norm, which was no report at all. Hamstrung by too little data too late, it was not any surprise that the FDA was often caught flat-footed when a problem was uncovered. There simply has not been adequate data input to make valid conclusions, greatly delaying our finding the key signals of a problem regarding a particular drug or device.


It makes good sense to have a pilot project that is linking multiple electronic databases, both public and private, for the purpose of post-approval drug surveillance. We need to be able to identify the most destructive side-effects or hazards of a drug or device as soon as possible, not years afterwards. However, the Sentinel Initiative will fall short if the main databases queried are only administrative databases, which contain the diagnoses used for billing, but not the richness of the medical record, whose delineation of co-morbid conditions and more detailed diagnoses are essential for correct conclusions. These will not be available to the surveillance officers unless they have access to clinical electronic databases, and those are not the main electronic databases that are to be queried. Without accurate clinical data, ambiguity and false conclusions will be more likely.


On the other hand, the most serious adverse events probably will be captured by the administrative database, as for example, an unexpected hospitalization, a sudden cessation of the medication following a new major diagnosis, etc. This alone is important and a great improvement over the present status quo.


In the future, it is highly likely that we will emerge with a deeper understanding of the risk-benefit profile of each of the drugs and devices we use. This will help us to use them more safely. Our focus should be to provide the right drug for the right patient, rather than to vainly search for one drug that is right for everyone. A more sophisticated, data driven strategy for therapeutics may emerge, but only if this promising effort is continued and expanded, not stifled by inadequate funding or over-reliance on the more convenient, but less accurate clinical data found in administrative databases.


We all agree that our patients deserve the most accurate system for post-approval drug safety. If the Sentinel Inititative is developed to include high quality clinical databases, then high quality, timely drug safety surveillance may finally be within our reach.





Welcome Editorial for the Patient Safety Exchange, May 8, 2008


2008-05-24 15:46:19

By: Richard Hellman, MD, FACP, FACE

Welcome to the AACE Patient Safety Exchange. This website represents our efforts to introduce you to the growing knowledge of the complexity of health care systems and the difficulty attaining and maintaining safer and higher quality care.


The website contains many separate parts that we hope will form a cohesive body of knowledge for the interested reader. The current “News” section will highlight key and timely issues regarding patient safety considerations of the practicing endocrinologist.


The item selected today is the recent news from the FDA regarding the significant number of injuries and deaths occurring in children and adolescents on insulin pumps. This is clearly neither a condemnation of the pumps themselves nor of the children and adolescents who use the pumps. The problem instead appears to be far more complex and subtle. Most of the children were trying hard to do the best they could, and many patients do extremely well on insulin pumps. In the right setting, with the right patients, the pumps are very safe. But the problem appears to be largely a result of a set of system problems, possibly a defect in human factor engineering, perhaps due to an inadequate or insufficient education and support, perhaps merely the wrong pump for the wrong patient.


Although without more data it is premature to conclude what the full scope of the problem is regarding both adult and pediatric insulin pump users, the question must be studied carefully, using the relevant science, if we are to decrease the number of injuries and fatal errors in our patients.


The “Glossary” of terms on this website will provide definitions of the terms important in patient safety, and examples to illustrate the meaning of the terms. The glossary is meant to be a minicourse, an introduction to the topic of patient safety. The library is comprehensive, with many of the “essential” references, and is a good introduction to the vast literature. It also covers many points of view, some that are controversial and provocative. We hope to soon have most of the references available for you as PDF files.


Please also look at the interactive portion of the website under the tab “Ask the Experts”, including the Physician’s Corner, the Allied Health Professional’s Corner, and the Patient’s Corner. The dialogue posted here, including both the questions and answers, touch on many subjects. Please send us your questions and any comments on what we have shared with you.


Most of us learn best by linking information to patient care. We have used case studies and accompanying commentaries for the purpose, and have been fortunate to have some excellent contributions in the areas of bone metabolism, thyroidology, and pituitary and hypothalamic disease, in addition to an example from the area of diabetes. We hope you will find these case studies useful, and the “Quick Take”, a brief case study on medication error, is both interesting and eye-opening.


The more formal presentations, the webcasts and podcasts, will continue to increase this year from just a very few at the inauguration of the website. We expect they will become a mainstay for the website. Please suggest topics and speakers. We want to be eclectic.


We will continue to translate as much of the website as possible into Spanish. In the future we hope to add other languages as well, dependent upon time and resources available.


Over the next year, we will add a CME component to the website. We also hope to have a section that will allow you to earn credit for maintenance of certification in the ABIM recertification program.


And please note the tab, “About Us”. We are proud of AACE, our parent organization, and very grateful for the fine work of all of our volunteer contributors and to the excellent support staff.


This has been a labor of love for us in developing this website. We hope you will also enjoy the information and find it interesting, provocative at times, but most of all helpful to you. Please send us your comments so we can fashion a website that will be enriched by your input.




Case Studies


Back to Basics

2008-05-02 14:29:37

By: Dr. Richard Hellman

The Case


A 48-year-old woman with insulin-dependent diabetes mellitus presents to the emergency department with right upper quadrant pain, fever, and leukocytosis, prompting admission for presumed cholangitis. Overnight, the patient was made NPO (nothing by mouth) in anticipation of an endoscopic retrograde cholangiopancreatography (ERCP) the following morning. The admitting medical team ordered an insulin sliding scale for the patient, and her blood glucose levels became very difficult to control in the ensuing hours. In the morning, the patient developed an anion gap and evidence of mild diabetic ketoacidosis. The physician evaluating the patient in the morning realized that no basal insulin was ordered and instituted a more appropriate regimen of insulin, and the patient underwent an uneventful ERCP and hospitalization.


The Commentary


This case is about a near miss, a tragic outcome narrowly averted by an alert physician who promptly corrected a series of earlier misjudgments that led to an inappropriate plan for insulin administration. The errant plan resulted in the development of diabetic ketoacidosis (DKA) in a seriously ill, infected patient who was NPO and awaiting an ERCP. If not recognized or anticipated, DKA carries the potential for lethal consequences.(1,2) A key error by the admitting team was not providing for a greatly increased basal insulin requirement, opting instead for a less effective insulin sliding scale algorithm. This decision making represented the wrong approach in this setting, an example of a rule-based error. Reason characterizes such situations as "strong but wrong" rule-based errors—the rule is strong in general but was inappropriate or misapplied in this situation.(3)


Metabolic stressors such as myocardial ischemia, surgery, cardiovascular collapse, and severe infections greatly increase insulin requirements. Both hyperglycemia and ketosis reduce immune defenses to infective agents.(4) During metabolic stress, the degree of ensuing insulin resistance is often hard to predict.(5) An insulin-dependent (Type 1) patient with an initial basal requirement of 0.6 units/hr may develop insulin requirements of 3–20 units/hr or more, rapidly resulting in marked hyperglycemia. This same patient will then develop severe insulin deficiency leading to ketoacidosis. Since the cardiovascular compensatory response to DKA requires significantly increased cardiac output, any underlying cardiac compromise will greatly increase the risk of death. Aggressive management of hyperglycemia in critically ill patients can greatly improve outcomes.(6) In this case, the admitting providers delivered an initial care plan that did not take into account the increased basal insulin requirements and the cascade of clinical events that may have ultimately led to an adverse outcome—one that was entirely preventable with better strategies for glycemic management.


Supplemental subcutaneous insulin algorithms alone, as ordered in this case, are ineffective and often inadequate to protect against the development of in-hospital DKA, which carries an increased risk of death.(2,7) A better strategy would involve initial assessment of insulin requirements and institution of a variable-rate insulin infusion with frequent glucose monitoring. An accompanying algorithm would still be required to adjust the insulin doses over time. The recommendation is to reach a goal and maintain glucose levels in the target range of 80–110 mg/dL during both the perioperative period and through the active infection.(2,5,8) During a variable-rate insulin infusion, any initial assumptions about insulin requirements are constantly revised based on both the actual glucose level and the magnitude of change during the observed time intervals. Barriers to the use of insulin infusions in acute care settings stem from the fact that, without key changes in the system of care, insulin infusion utilization is fraught with difficulty.(1)


System Changes to Improve Glycemic Control


A well-studied and well-designed, uniform algorithm for insulin administration should be implemented in all acute care settings. All providers (e.g., nurses, physicians, and clinical pharmacists) who care for diabetic patients need to be familiar with the details of the algorithm. Although many excellent insulin algorithms exist in the public domain, the algorithm alone is seldom enough to improve patient outcomes.(2,5,8) Education and focused training for all providers are essential, since those experienced in using a specific insulin algorithm do so more effectively than those who understand neither the context nor the orders.(1,2,9) The Figure shows an example algorithm; these orders have been field tested over many years in six major hospitals and have been used safely in diverse clinical settings.


Another essential aspect of implementing an insulin algorithm is relying on interdisciplinary involvement. Many efforts to improve glycemic control focus on either nurses or physicians when in reality, a patient relies on both to safely order, monitor, and administer appropriate amounts of insulin. Often, a lack of improvement in glycemic control when using a well-designed algorithm may be due to an important new or previously unnoted clinical problem—perhaps a nursing error in the use of the algorithm, an incorrect insulin amount in the delivery system, or an increase in insulin resistance from early cardiovascular collapse or sepsis (with inappropriate adjustments to insulin delivery). Since timely physician input may be life-saving, the algorithm must include criteria as to when the physician must be consulted.


Insulin infusions can be used safely throughout the hospital (i.e., they do not require admission to a step-down unit or intensive care unit) when the system of care allows for adequate training and supervision of those using the selected algorithms. Moreover, the use of hospital electronic health records (EHR) can make bedside glucose results available in real time throughout the hospital system and in remote locations, allowing timely oversight.(10) Computerized physician order entry systems (CPOE) can also be used with approved insulin algorithms to reduce errors in the implementation of medical orders.(10,11)


The Pitfalls of Sliding Scale Insulin Use


In closing, we should also note why the sliding scale–only approach led to near disaster in this patient. Nearly all of the published sliding scale approaches give inadequate amounts of insulin subcutaneously.(12) Subcutaneous absorption may be slow and erratic, particularly in critical illness, and when insulin is given subcutaneously every 4–6 hours, the insulin dosage is often "too little too late." Also, in many sliding scale algorithms, if the glucose level is under 150 mg/dL, there may be no insulin recommended at all, a potentially disastrous error in an insulin-dependent patient with high basal requirements. Insulin-dependent patients develop uncontrolled ketone production, even if fasting, within minutes of the onset of severe insulin deficiency. Dr. Stephen Clement presented two cases of patients with in-hospital deaths due to cardiovascular complications of new-onset ketoacidosis after the use of sliding scale insulin algorithms overnight.(13) Comments also have been published regarding catastrophic insulin errors stemming from inappropriate treatment approaches.(1,2)


Take-Home Points


    • Metabolic stressors, such as infections, greatly increase basal insulin requirements, putting patients at risk for hyperglycemia and DKA.
    • Sliding scale insulin algorithms alone are not useful for achieving normoglycemia, which is the standard of care for critically ill patients with diabetes.
    • Variable-rate insulin infusions are extremely useful, but often underutilized, in treating hyperglycemia associated with severe infections. Implementing these algorithms requires adequate training and supervision of staff and close monitoring of patients.
    • Information technology (e.g., EHR and CPOE) can serve as a useful tool to improve monitoring and ordering of insulin, resulting in fewer errors.



Source: Richard Hellman, MD, Clinical Professor of Medicine, University of Missouri-Kansas City School of Medicine. Case & Commentary: Medicine. Back to Basics. Morbidity & Mortality Rounds on the Web. Agency for Healthcare Research and Quality, Rockville, MD. March 2007. Used with permission.


Richard Hellman, MD

Clinical Professor of Medicine

University of Missouri-Kansas City School of Medicine




Figure. Example Insulin Protocol. (Reprinted with permission of Richard Hellman, MD.)


Click to view pdf

    1. Hellman R. A systems approach to reducing errors in insulin therapy in the inpatient setting. Endocr Pract. 2004;10(suppl 2):100-108. [go to PubMed]
    2. Hellman R. Strategies to reduce medical errors in the management of diabetes. Updates: Harrison's Internal Medicine [AccessMedicine Web site]. October 9, 2002. Available at: Accessed March 15, 2007.
    3. Reason J. Human Error. Cambridge, UK: Cambridge University Press; 1990.
    4. McManus LM, Bloodworth RC, Prihoda TJ, Blodgett JL, Pinckard RN. Agonist-dependent failure of neutrophil function in diabetes correlates with extent of hyperglycemia. J Leuk Biol. 2001;70:395-404. [go to PubMed]
    5. van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin therapy in the critically ill patients. N Engl J Med. 2001;345:1359-1367. [go to PubMed]
    6. Van den Berghe G, Wouters PJ, Bouillon R, et al. Outcome benefit of intensive insulin therapy in the critically ill: insulin dose versus glycemic control. Crit Care Med. 2003;31:359-366. [go to PubMed]
    7. Trence DL, Kelly JL, Hirsch IB. The rationale and management of hyperglycemia for in-patients with cardiovascular disease: time for change. J Clin Endocrinol Metab. 2003;88:2430-2437. [go to PubMed]
    8. Goldberg PA, Siegel MD, Sherwin RS, et al. Implementation of a safe and effective insulin infusion protocol in a medical intensive care unit. Diabetes Care. 2004;27:461-467. [go to PubMed]
    9. Hellman R. Patient safety and inpatient glycemic control: translating concepts into action. Endocr Pract. 2006;12(suppl 3):49-55. [go to PubMed]
    10. Bates D, Clark NG, Cook RI, Hellman R, et al. American College of Endocrinology and American Association of Clinical Endocrinologists position statement on patient safety and medical system errors in diabetes and endocrinology. Endocr Pract. 2005;11:197-202. [go to PubMed]
    11. Bates DW, Gawande AA. Improving safety with information technology. N Engl J Med. 2003;348:2526-2534. [go to PubMed]
    12. Clement S, Braithwaite SS, Magee MF, et al, for the Diabetes in Hospitals Writing Committee. Management of diabetes and hyperglycemia in hospitals. Diabetes Care. 2004;27:553-591. [go to PubMed]
    13. Clement S. Inpatient diabetes care and insulin delivery. Presented at: Patient Safety and Medical System Errors in Diabetes and Endocrinology Consensus Conference; January 9, 2005; Washington, DC.