Practice Uncertainties

Should inpatient hyperglycaemia be treated?

BMJ 2013; 346 doi: (Published 17 January 2013) Cite this as: BMJ 2013;346:f134
  1. Ketan Dhatariya, consultant in diabetes and endocrinology
  1. 1Elsie Bertram Diabetes Centre, Norfolk and Norwich University Hospitals NHS Foundation Trust, Norwich NR4 7UY, UK
  1. Correspondence to: K Dhatariya ketan.dhatariya{at}
  • Accepted 20 December 2012

Two large scale randomised controlled trials in the 1990s were the first such trials to show that the control of blood glucose helped to prevent long term complications in people with types 1 and 2 diabetes.1 2 Glucose concentrations can rise not only in people with pre-existing diabetes, but also, for short periods, in people without the condition—in particular, during times of acute illness, when it is called stress hyperglycaemia.3 As discussed below, data show that raised blood glucose concentrations in people with and without a previous diagnosis of diabetes are associated with short term harm. However, whereas the benefits of good glycaemic control over a long period in people with diabetes are well established, uncertainty remains about whether treating transient hyperglycaemia, in particular in hospital inpatients, makes any difference to short term outcomes.

Sometimes the question is moot and treating hyperglycaemia is crucial—for example, in diabetic ketoacidosis or hyperosmolar hyperglycaemic states4 or in patients who are symptomatic from their hyperglycaemia. This article does not discuss these states.

What is the evidence of the uncertainty?

I searched PubMed, the Cochrane Library, and Clinical Evidence to identify publications that concerned the outcomes of hyperglycaemia in hospital inpatients and those that dealt with treating the hyperglycaemia. Since the two trials in the 1990s,1 2 other studies have also shown that hyperglycaemia in inpatients with and without pre-existing diabetes is associated with poor outcomes. However, most trials were observational, with only a few randomised controlled trials. A meta-analysis of 34 randomised control trials assessing perioperative insulin infusion in 2192 surgical patients concluded that “perioperative insulin infusion may reduce mortality but increases hypoglycaemia in patients who are undergoing surgery.”5 However, only 14 of these studies included patients with diabetes, with 13 studies excluding them and the rest not reporting whether patients with diabetes were included.

Observational data from an unselected cohort of over 1500 acute general medical admissions with and without diabetes showed that length of stay, readmission rates, and 30 day mortality rates rose with higher blood glucose concentrations.6 Other observational evidence from hospital episode statistics based on discharge coding of over four million patients showed that those who also had diabetes stayed in hospital the longest, regardless of the specialty.7 There is also a wealth of observational data to show that elective or emergency surgical patients with or without pre-existing diabetes also have poorer outcomes when they have high preoperative glycated haemoglobin concentrations (reflecting poor preadmission glycaemic control) or high perioperative blood glucose concentrations.8 9

People with stress hyperglycaemia may be at risk of developing type 2 diabetes in the long term. However, evidence from intervention studies is sparse or conflicting on whether aggressive treatment of the hyperglycaemia during a patient’s hospital stay makes a difference to short or long term outcomes or even affects outcomes related to their cause for admission. Indeed, data from well conducted large randomised controlled trials and observational studies show that the use of glucose lowering agents—in particular, insulin—are associated with increased levels of harm, in the form of severe hypoglycaemia.10 11

A few randomised controlled trials show that short term, tight glycaemic control using insulin therapy in intensive care seemed to reduce mortality, infection rate, and length of hospital stay.12 13 Other well conducted randomised controlled trials in intensive care patients have been either equivocal14 15 or associated with harm, with the largest such study of over 6000 patients showing that tight glycaemic control was associated with higher incidence of severe hypoglycaemia and increased mortality.16 Randomised controlled trials have shown that short term tight glycaemic control can also help patients who have cardiac surgery—benefits included fewer sternal wound infections.5 However, patients in intensive care or having cardiac surgery are a minority. Furthermore, the data from randomised controlled trials for patients presenting with acute coronary syndromes remain conflicting,17 18 although this is probably because of poor study design and recruitment. The data for acute coronary syndrome seem so contradictory that the American Heart Association avoided the topic of hyperglycaemia in its 2008 position paper on the management of acute coronary syndrome, despite a substantial proportion of patients presenting with concurrent hyperglycaemia, and hyperglycaemia being associated with poor outcomes.19

There are good theoretical reasons why glucose reduction with insulin should be beneficial, with reductions in endothelial dysfunction, immune dysfunction, and the maintenance of adequate vasodilatation.20 But insulin use in any patient with hyperglycaemia is fraught with problems and is often used incorrectly or ineffectively—the use of subcutaneous “sliding scales” being one such problem.21 Precipitating severe hypoglycaemia by aggressive glucose lowering with insulin is a major concern, as is the lack of confidence among junior doctors in managing the condition.22 However, recently published documents and education packages (available at have sought to reduce these errors.23 Uncertainty also remains about the glucose targets that should be aimed for and the best agents to achieve these.

The data presented show that high glucose concentration in people with and without diabetes is associated with poor outcomes. However, as I found no directly relevant systematic reviews it remains to be determined if the raised blood glucose is the cause of the poor outcomes or if it is just an epiphenomenon.

Is ongoing research likely to provide relevant evidence?

Large, well conducted randomised controlled trials are needed in several patient populations to establish whether glycaemic control reduces or prevents the harms associated with hyperglycaemia.

All of the factors discussed here mean that very large numbers of patients would be needed over many sites for a long time. A search on shows that several studies of glycaemic control in hospital inpatients are ongoing, but most are studying small numbers of patients in specialised populations. Although these smaller scale studies help greatly, it is important that the methods used in these smaller studies are adequate to allow rigorous and meaningful meta-analyses to be conducted to help resolve the uncertainties raised in this article. Definitive, large studies are likely to be very expensive, and therefore in the current economic environment are unlikely to be conducted in the size needed to answer such questions.

What should we do in light of the uncertainty?

Given the data showing that hyperglycaemia in hospital inpatients is detrimental, all adult patients with or without a pre-existing diagnosis of diabetes should have their blood glucose measured on admission. If they are found to be hyperglycaemic then efforts should be made to control their glucose concentrations on the basis of pragmatic consensus documents drawing largely on the best available observational data previously described. For example, the guidelines commissioned by England’s NHS Diabetes on the perioperative management of patients with diabetes having surgery recommend that for inpatients needing a prolonged starvation time (that is, more than one missed meal) a variable rate intravenous insulin infusion should be used, with the aim of keeping their blood glucose concentrations ideally between 6 mmol/L and 10 mmol/L (with a range of 4 mmol/L to 12 mmol/L being considered “acceptable”).24 For those whose diabetes status is not known, no accepted guidelines exist; however, the recommendation is to have the same glycaemic targets as for people with diabetes.3

Recommendation for further research

  • All adult hospital inpatients, including elective and emergency patients, from all medical and surgical specialties, and with or without a previous diagnosis of diabetes

Intervention and comparisons
  • Initial phase: observational data to assess the relation between blood glucose and glycated haemoglobin concentrations on admission, and outcomes

  • Next phase: randomised controlled trials to compare good glycaemic control (such as target blood glucose concentrations of 4-12 mmol/L) with usual standard of care, taking into account multiple confounders in these groups, such as age, comorbidities, pre-existing diabetes, medication use

  • To include in-hospital mortality, 30 day mortality, length of stay, 30 day hospital readmission rates, and postoperative complication rates for surgical patients. Other outcomes would depend on the specialty.


Cite this as: BMJ 2013;346:f134


  • This is one of a series of occasional articles that highlight areas of practice where management lacks convincing supporting evidence. The series adviser is David Tovey, editor in chief, the Cochrane Library. To suggest a topic for this series, please email us at uncertainties{at}

  • Contributors: KD is the sole author and contributor.

  • Competing interests: The author has completed the ICMJE uniform disclosure form at (available on request from the corresponding author) and declares: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years. KD is an author on several of the Joint British Diabetes Society Inpatient Care Group guidelines for the management of various aspects of inpatient diabetes care, and these activities could appear to have influenced the submitted work.

  • Provenance and peer review: Not commissioned; externally peer reviewed.