Absent such evidence let's start with what we know. We know we don't want patients in the hospital to slip into DKA or a hyperosmolar state, or to develop fluid and electrolyte disturbances as a result of marked hyperglycemia. On the other extreme we don't want them hypoglycemic. Is anything and everything in between those extremes OK? Probably not, but without evidence based targets what should we do? We can do little more than arbitrarily choose regimens based on what seems “reasonable” according to clinical expertise.
For critically ill patients an insulin drip affords flexibility of titration and reliability of drug delivery as opposed to the erratic sub Q absorption that might occur if the patient is in shock. I have posted extensively on glycemic control in critical illness in previous posts    and will not elaborate further here.
For the non critically ill setting we have some data on surgical patients:
RESEARCH DESIGN AND METHODS This randomized multicenter trial compared the safety and efficacy of a basal-bolus insulin regimen with glargine once daily and glulisine before meals (n = 104) to sliding scale regular insulin (SSI) four times daily (n = 107) in patients with type 2 diabetes mellitus undergoing general surgery. Outcomes included differences in daily blood glucose (BG) and a composite of postoperative complications including wound infection, pneumonia, bacteremia, and respiratory and acute renal failure.
RESULTS The mean daily glucose concentration after the 1st day of basal-bolus insulin and SSI was 145 ± 32 mg/dL and 172 ± 47 mg/dL, respectively (P less than 0.01). Glucose readings less than140 mg/dL were recorded in 55% of patients in basal-bolus and 31% in the SSI group (P less than 0.001). There were reductions with basal-bolus as compared with SSI in the composite outcome [24.3 and 8.6%; odds ratio 3.39 (95% CI 1.50–7.65); P = 0.003]. Glucose less than 70 mg/dL was reported in 23.1% of patients in the basal-bolus group and 4.7% in the SSI group (P less than 0.001), but there were no significant differences in the frequency of BG less than 40 mg/dL between groups (P = 0.057).
CONCLUSIONS Basal-bolus treatment with glargine once daily plus glulisine before meals improved glycemic control and reduced hospital complications compared with SSI in general surgery patients. Our study indicates that a basal-bolus insulin regimen is preferred over SSI in the hospital management of general surgery patients with type 2 diabetes.
This study only compared two insulin strategies. It did not evaluate particular glycemic targets. A similar study on non sugical ward patients was done by the same investigators but it addressed only the surrogate outcome of glycemic control, not whether better control was associated with improvement in meaningful clinical outcomes.
Both studies showed that a basal-bolus regimen provided tighter glycemic control than did a simple sliding scale when the latter was used as the sole modality.
So for ward patients we can summarize it this way:
Optimal glycemic targets are unknown.
Basal-bolus regimens provide better glycemic control than sliding scale (SSI) as the sole modality.
Basal-bolus regimens are associated with better clinical outcomes than SSI in surgical patients.
It is not known whether basal-bolus insulin provides better clinical outcomes in medical patients.
With next to no evidence to go on the guidelines are based on physiologic rationale and expert opinion. The review cites the guidelines and proposes an algorithm.
A few practical points from the review:
Sliding scale coverage is discouraged by the authors but may be acceptable for short hospital stays in which it is anticipated that the patient will be NPO for a significant period of time.
What to do with oral agents is a matter for clinical judgment but the authors offer some cautions:
We tend to avoid the use of oral antihyperglycemic agents. Sulfonylureas may lead to hypoglycemia if nutrition is interrupted. There is no inherent reason that metformin could not be used, but given its risk for lactic acidosis and the frequent complicating features of hospitalized patients that might increase this risk (renal failure, dehydration, severe heart failure, acidosis), it would appear best to avoid this medication. Thiazolidinediones can be continued if there are no concerns of heart failure, although an interruption in therapy of a few days should not appreciably alter glucose levels. Incretin-based therapies (DPP-4 inhibitors and GLP-1 receptor agonists) are ideally used in patients who are eating because they target mainly postprandial glucose and so should probably be avoided when nutritional intake is forbidden or tentative. GLP-1 receptor agonists may also result in nausea so are best avoided in this setting. Some experts have proposed that, because these drug classes are not associated with hypoglycemia and because their activity appears to be in part related to the degree of hyperglycemia, they may have a role in acutely ill patients. Adequately powered, randomized clinical trials will be needed.
Finally the authors address discharge planning and emphasize the need for preparations to begin 2-3 days in advance if possible. Transitioning to post-hospital diabetes care is difficult for the hospitalist, who will not be managing the patient as an outpatient.
HT to Hospital Medicine Virtual Journal Club.