A Systematic Review Supporting the Endocrine Society Clinical Practice Guideline for the Management of Hyperglycemia in Adults Hospitalized for Noncritical Illness or Undergoing Elective Surgical Procedures

Abstract

Context

Individuals with diabetes or newly recognized hyperglycemia account for over 30% of noncritically ill hospitalized patients. Management of hyperglycemia in these patients is challenging.

Objective

To support development of the Endocrine Society Clinical Practice Guideline for management of hyperglycemia in adults hospitalized for noncritical illness or undergoing elective surgical procedures.

Methods

We searched several databases for studies addressing 10 questions provided by a guideline panel from the Endocrine Society. Meta-analysis was conducted when feasible. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology was used to assess certainty of evidence.

Results

We included 94 studies reporting on 135 553 patients. Compared with capillary blood glucose, continuous glucose monitoring increased the number of patients identified with hypoglycemia and decreased mean daily blood glucose (BG) (very low certainty). Data on continuation of insulin pump therapy in hospitalized adults were sparse. In hospitalized patients receiving glucocorticoids, combination neutral protamine hagedorn (NPH) and basal-bolus insulin was associated with lower mean BG compared to basal-bolus insulin alone (very low certainty). Data on NPH insulin vs basal-bolus insulin in hospitalized adults receiving enteral nutrition were inconclusive. Inpatient diabetes education was associated with lower HbA1c at 3 and 6 months after discharge (moderate certainty) and reduced hospital readmissions (very low certainty). Preoperative HbA1c level < 7% was associated with shorter length of stay, lower postoperative BG and a lower number of neurological complications and infections, but a higher number of reoperations (very low certainty). Treatment with glucagon-like peptide-1 agonists or dipeptidyl peptidase-4 inhibitors in hospitalized patients with type 2 diabetes and mild hyperglycemia was associated with lower frequency of hypoglycemic events than insulin therapy (low certainty). Caloric oral fluids before surgery in adults with diabetes undergoing surgical procedures did not affect outcomes (very low certainty). Counting carbohydrates for prandial insulin dosing did not affect outcomes (very low certainty). Compared with scheduled insulin (basal-bolus or basal insulin + correctional insulin), correctional insulin was associated with higher mean daily BG and fewer hypoglycemic events (low certainty).

Conclusion

The certainty of evidence supporting many hyperglycemia management decisions is low, emphasizing importance of shared decision-making and consideration of other decisional factors.

Individuals with diabetes account for over 30% of noncritically ill hospitalized patients (1, 2). Another 12% to 25% of hospitalized patients without a known prior history of diabetes will experience hyperglycemia during their inpatient stay (1, 3, 4). Inpatient hyperglycemia, defined as any blood glucose (BG) ≥ 140 mg/dL, is associated with an increase in the risk for adverse clinical outcomes, including infections, prolonged hospital length of stay, post-discharge disability, and mortality (4-6). The 2012 Endocrine Society Clinical Guideline “Management of Hyperglycemia in Hospitalized Patients in Non-Critical Care Settings” recommended that BG levels be maintained between 100 and 180 mg/dL for the majority of patients with diabetes or newly recognized hyperglycemia to minimize risk for these adverse outcomes (7, 8). These earlier recommendations were derived from a limited number of studies conducted in noncritical care settings or translated from studies performed in critical care settings (9-13).

Since 2012, there have been important advances in the number of pharmacologic agents for treating diabetes in the outpatient setting, as well as advances in available technologies for monitoring therapy (14-16). Several important studies have been published in the past decade that address issues for which there was little data to support recommendations provided in the earlier guideline. Examples of these advances include the introduction of basal insulin preparations with more consistent pharmacokinetic profiles, the use of neutral protamine hagedorn (NPH, also known as isophane)-based insulin regimens for treating glucocorticoid- or enteral nutrition–associated hyperglycemia, the role of inpatient diabetes education on post-discharge outcomes, and the emergence of continuous glucose monitoring (CGM) devices that demonstrate improved glycemic measures with less hypoglycemia in both nonhospitalized and hospitalized patients. These advances provide the opportunity to facilitate goal-directed glycemic management in the hospital while also avoiding hypoglycemia.

In addition to these advances in glycemic management, new questions have emerged for which there is debate but little data supporting one method over another, resulting in variability among institutions regarding these practices. These areas of debate include the use of noninsulin therapies in the hospital, the administration of preoperative carbohydrate-containing beverages to improve postoperative outcomes, and recommendations for preoperative glycemic control prior to elective surgical procedures. A longstanding and ongoing debate in the management of inpatient hyperglycemia relates to the practice of using sliding scale insulin (SSI), referred to in this report as correction insulin, in hospitalized patients.

The Endocrine Society identified management of hyperglycemia in hospitalized patients in noncritical care settings as a topic for review and update of the 2012 guideline. The writing panel identified several decision-making dilemmas faced by health care providers who provide care for these patients. To support the guideline development process and summarize the evidence base, we conducted this systematic review and meta-analysis.

Methods

This reporting of the review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statements (17). The study protocol and the specific questions of this review were developed by a guideline panel from the Endocrine Society.

Eligibility Criteria

The panel members prioritized 10 questions addressing hyperglycemia in patients hospitalized for a noncritical illness or undergoing elective surgical procedures. The 10 questions are presented using the PICO format (patient, intervention, comparison, outcomes) along with the eligibility criteria in eTable 1 in the supplement (18).

Data Sources and Search Strategies

Comprehensive searches of several databases from each database’s inception to December 8, 2021, any language, were conducted. The databases included Ovid MEDLINE(R) and Epub Ahead of Print, In-Process & Other Non-Indexed Citations, and Daily, Ovid EMBASE, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews, and Scopus. Searches for the 10 questions were conducted through 8 search strategies designed and conducted by a medical reference librarian with input from the investigators. Additional references identified by the guideline committee were considered. The actual strategies listing all search terms used and how they were combined is provided in eTable 2 in the supplement (18).

Study Selection

Pairs of independent reviewers screened titles and abstracts for eligibility. Studies included by either reviewer were retrieved for full-text screening. Two independent reviewers screened the full texts; any disagreements between the 2 reviewers were resolved by a third reviewer.

Data Extraction

The data from the included studies were extracted by pairs of independent reviewers. We extracted data on patient characteristics (number of patients, mean age, female percentage, and details of intervention) and outcomes of interest (mean daily BG, % of BG > 180 and 250 mg/dL or % time in these ranges, hypoglycemia (BG < 54 mg/dL and < 70 mg/dL), hospital length of stay (LOS), nurse time and effort, nurse satisfaction, and patient satisfaction.

Methodologic Quality and Certainty of the Evidence

The risk of bias in the included studies was assessed using the Newcastle-Ottawa scale for nonrandomized studies (19), and the Cochrane risk of bias tool for randomized controlled trials (RCTs) (20). The risk of bias was assessed at the study level by 2 independent reviewers and disagreements were resolved by discussion and consensus. The Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach was used to rate the certainty of evidence for each outcome (21).

Data Synthesis and Analysis

For comparative studies, relative risk (RR) for binary outcomes were calculated, and the 95% CI was estimated. For continuous outcomes, the mean difference (MD) was calculated. The incidence rate ratio (ie, the ratio of the incidence rate of events within a given period between the intervention and the comparison) was calculated for adverse events. Due to anticipated heterogeneity, meta-analysis using the random-effects model was used. The I² index and Cochrane Q test were used to evaluate heterogeneity. All statistical analyses were conducted using R version 4.1.3 (R Foundation for Statistical Computing). When meta-analysis was not feasible, data were presented narratively.

Results

Study Selection

The literature searches identified 7482 citations, of which we finally included 94 studies. The study selection process is depicted in eFigure 1 in the supplement (18). The characteristics of included studies, details of the interventions reported in each study, and the risk of bias assessment tables are provided in eTables 3 and 4 in the supplement (18), respectively. The majority of patients represented in the studies were identified as having type 2 diabetes, with only a small number of studies indicating that patients with type 1 diabetes were included. Several studies did not identify the type of diabetes for the included subjects.

Question 1. Continuous Glucose Monitoring vs Bedside Capillary Blood Glucose Testing in Adults With Diabetes Hospitalized for Noncritical Illness

Nine studies (5 RCTs (22-26), 4 non-RCTs (27-30)) with 536 patients were included. Two studies had a low risk of bias; 7 studies had an overall moderate to high risk of bias (eTable 3.1 in the supplement (18)). Compared with capillary blood glucose, CGM was associated with lower mean daily blood glucose levels (8 studies, MD = −9.00 mg/dL [95% CI, −15.20, −2.79], moderate certainty) and a higher proportion of patients detected with hypoglycemia (BG < 54 mg/dL) (2 studies, RR = 4.67 [95% CI, 1.05, 20.74], very low certainty).

There was no statistically significant difference in the percentage time with blood glucose > 180 and 250 mg/dL, or the detection rate of hypoglycemic events (low/very low certainty). The results are summarized in eTable 5.1, and in eFigure 2.1 in the supplement (18).

Question 2. NPH Insulin vs Basal-Bolus Insulin in Adults With Hyperglycemia Receiving High Dose Glucocorticoids Hospitalized for Noncritical Illness

Seven studies (6 RCTs (24, 31-35) 1 non-RCT (36)) with 429 patients were included. One study had low risk of bias. The rest of the studies had an overall moderate to high risk of bias (eTable 3.2 in the supplement (18)). The included studies compared NPH insulin in one arm (with or without another insulin) to several insulin regimes (basal, basal-bolus, NPH + basal, and NPH + basal-bolus insulin). Addition of NPH to home insulin therapy (mix of basal-bolus and basal only) was associated with lower mean daily BG compared with home insulin therapy alone (1 study, mean = −42.45 mg/dL [95% CI, −63.21, −21.69], moderate certainty).

Combination of NPH + basal-bolus insulin was associated with lower mean daily BG when compared with basal-bolus therapy (3 studies, mean = −40.6 [95% CI, −75.14, −6.06], very low certainty). There was no statistically significant difference in the rates of hypoglycemia or hospital LOS, very low certainty). The results are summarized in eTable 5.2, and in eFigure 2.2 in the supplement (18).

Question 3. Continuation of Insulin Pump Therapy in Adults Hospitalized for Noncritical Illness

Two non-RCT studies (37, 38) with 186 patients were included. The overall risk of bias was moderate (eTable 3.3 in the supplement (18)). A study by Kannan et al (38) compared 3 groups: no education combined with continuous subcutaneous insulin infusion (CSII) (29 patients), education combined with CSII (15 patients), and no education combined with multiple daily insulin injections (MDI) (6 patients). The study showed no statistically significant difference among the 3 groups in terms of the frequency of hyperglycemia and hypoglycemic events and the mean daily blood glucose (very low certainty).

Another study by Cook et al (37) included 136 patients with 253 hospitalizations. Glycemic measures were examined in 164 (65%) patients while using CSII (“‘pump on”), 50 (20%) using CSII intermittently, and 38 (15 %) where CSII was discontinued (“pump off”). There were no significant differences in the number of episodes of hypoglycemia in patients who continued pump use vs those who changed to Subcutaneous injection insulin. Yet, episodes of severe hypoglycemia (<40 mg/dL) and hyperglycemia were significantly less common among pump users (very low certainty). The results are summarized in eTable 5.3, and in eFigure 2.3 in the supplement (18).

Question 4. Inpatient Diabetes Education for Adult Patients With Diabetes Hospitalized for Noncritical Illness

Ten studies (4 RCTs (39-42), 6 non RCTs (43-48)) with 3809 patients were included. The overall risk of bias was moderate to high (eTable 3.4 in the supplement (18)).

Inpatient diabetes education was associated with a significant reduction in glycated hemoglobin (HbA1c) at 3 months and 6 months after discharge (4 studies, MD = −1.27% [95% CI, −1.95, −0.59]) and (3 studies, MD = −1.09% [95% CI, −1.67, −0.50]), respectively. Certainty of evidence was moderate. Inpatient diabetes education was also associated with a significant reduction in hospital readmission (RR = 0.72 [95% CI, 0.90, 0.87]). Certainty of evidence was very low.

There was no statistically significant difference in hospital LOS or patient satisfaction. Certainty of evidence was very low. The results are summarized in eTable 5.4, and in eFigure 2.4 in the supplement (18).

Question 5. Prespecified Preoperative HbA1c or Blood Glucose Testing in Adults With Diabetes Undergoing Elective Surgical Procedures

Forty-four non-RCT studies (49-92) with 127 791 patients were included. The overall risk of bias was moderate to high (12 studies with low risk of bias, 14 with moderate risk of bias, and 18 with high risk of bias; eTable 3.5 in the supplement (18)). Preoperative HbA1c was assessed based on 3 threshold levels (7%, 7%-8%, and > 8%).

An HbA1c level of less than 7% was associated with shorter length of stay (11 studies, MD = −0.45 [95% CI, −0.89, 0], very low certainty), lower mean postoperative BG levels (4 studies, MD = −46.48 mg/dL [95% CI], −76.67, −16.3, very low certainty), and a lower number of neurological complications and infections (11 studies, incidence rate ratio = 0.51 [95% CI, 0.37, 0.70], very low certainty) and (10 studies, odds ratio = 0.54; 0.40, 0.73, very low certainty), respectively. However, an HbA1c level of less than 7% was associated with a higher number of reoperations (6 studies, incidence rate ratio = 1.40 [95% CI, 1.03, 1.92], very low certainty). There was no significant difference in the rate of cardiac, renal, and respiratory complications (very low certainty).

Prespecified preoperative HbA1c level (7%-8%, or 8%) were not associated with significant changes in the length of hospital stay, mean blood glucose level, hypoglycemia, hyperglycemia, infection, or adverse events (very low certainty). The results of the meta-analyses of outcomes and adverse events are summarized in in eTable 5.4, and in eFigure 2.4 in the supplement (18). Detailed summaries of studies without sufficient data for meta-analysis are shown in eTable 6 in the supplement (18).

Question 6. NPH Insulin vs Basal-Bolus Insulin in Adult Patients Hospitalized for Noncritical Illness Receiving Enteral Nutrition With Diabetes-Specific and Nonspecific Formulations

Four studies (1 RCT (13), 3 non-RCTs (93-95)) with 171 patients were identified. One RCT had a moderate risk of bias, 2 cohort studies had a low risk of bias, and 1 cohort study had a high risk of bias (eTable 3.6 in the supplement (18)). A comparison of different insulin regimens in the management of hyperglycemia in patients receiving bolus or nocturnal tube feeds revealed no significant differences in hospital LOS, hypoglycemia, mean daily BG, or mortality. Certainty of evidence was very low. The results are summarized in in eTable 5.6, and in eFigure 2.6 in the supplement (18).

Question 7. Insulin vs Noninsulin Therapies in Adults With Hyperglycemia (Including Those With and Without Known Diabetes) Hospitalized for Noncritical Illness

Five RCTs (96-100) with 646 patients were identified. All the included studies had a high risk of bias (eTable 3.7 in the supplement (18)). The studies reported that treatment with GLP-1 agonists or dipeptidyl peptidase-4 (DPP-4) inhibitors in patients with type 2 diabetes hospitalized in noncritical settings was associated with a lower frequency of hypoglycemic events than insulin therapy (2 studies, RR = 0.09 [95% CI, 0.01, 0.66], low certainty) and (3 studies, RR = 0.27 [95% CI, 0.09, 0.84], low certainty), respectively. There was no significant difference in the hospital LOS between GLP-1 agonist therapy when compared with insulin (very low certainty). DDP-4 inhibitor use was associated with statistically significant longer LOS when compared with insulin (3 studies, MD = 0.94 [95% CI, 0.51, 1.36], moderate certainty). There was no statistically significant difference between insulin and noninsulin therapies in the mean daily BG or in the percentage time of hyperglycemia (> 180 mg/dL and > 250 mg/dL, low/very low certainty). Adverse events and complications were rare and not statistically different (acute renal failure, respiratory failure, cardiovascular complications, and infection, very low certainty), but there was significantly more nausea/vomiting with GLP-1 agonists compared with insulin (2 studies, RR = 5.95 [95% CI, 1.07, 33.03], low certainty). The results of meta-analyses of the outcomes, adverse events are summarized in eTable 5.7, and in eFigure 2.7 in the supplement (18).

Question 8. Caloric Oral Fluids Before Surgery in Adults With Diabetes Undergoing Surgical Procedures

Three studies (1 RCT (101), 2 non-RCTs (102, 103)) with 357 patients were included. The overall risk of bias was moderate in all the studies (eTable 3.8 in the supplement (18)). There was no significant difference in the number of hypoglycemic events, mean daily BG, hospital LOS, and patient satisfaction when comparing patients who did or did not receive preoperative caloric oral fluid. Certainty of evidence was very low. The results are summarized in in eTable 5.8, and in eFigure 2.8 in the supplement (18).

Question 9. Carbohydrate Counting in Adults With Diabetes Hospitalized for Noncritical Illness

Three studies (1 RCT (104), 2 non-RCTs (105, 106)) with 415 patients were identified. All had high risk of bias (eTable 3.9 in the supplement (18)). There was no significant difference in the number of hypoglycemic events between adults with diabetes hospitalized for noncritical illness who counted carbohydrates for prandial insulin dosing and those who did not. Counting carbohydrates for prandial insulin dosing was comparable with no carbohydrate counting in terms of patient satisfaction, hospital LOS, and mean daily BG. Certainty of evidence was very low. The results are summarized in eTable 5.9, and in eFigure 2.9 in the supplement (18).

Question 10. Sliding Scale Insulin (Correctional Insulin) in Adults With Hyperglycemia Hospitalized for Noncritical Illness

Nine studies (6 RCTs (11-13, 107-109), 3 non-RCTs (110-112)) with 1213 patients were included. The overall risk of bias was high in 7 studies, moderate in 1 study and low in 1 study (eTable 3.3 in the supplement (18)).

Compared with scheduled insulin (basal-bolus insulin + correctional insulin or basal insulin + correctional insulin), correctional insulin used alone was associated with higher mean daily BG when compared with basal-bolus insulin (7 studies, MD = 25.82 [95% CI, 15.52, 36.13], low certainty). Correctional insulin was associated with fewer hypoglycemic events and a lower risk of hypoglycemia (low certainty). There was no significant difference in the length of hospital stay. The results are shown in eTable 5.10, and in eFigure 2.10 in the supplement (18).

Discussion

The care of hospitalized patients with diabetes and/ or hyperglycemia continues to be an area of interest and concern for patients as well as for hospital systems. Implementation of protocols and hiring of personnel with expertise in guiding safe and effective glycemic management and education can be seen as time, labor, and cost intensive by many hospitals. The 2012 Endocrine Society Clinical Guideline focused on defining glycemic targets along with suggested strategies for achieving a BG range of 100 to 180 mg/dL (8). These targets and management strategies were extrapolated from large studies in critically ill patient populations or smaller studies in noncritically ill patients (6, 12, 13, 113). The 2012 Clinical Guideline and accompanying systematic review described a need for RCTs performed in noncritical care settings to facilitate decision-making in this population (114). Although no RCT has yet been conducted to address optimal glycemic targets for noncritically ill hospitalized patients, the literature addressing specific approaches for effective and safe inpatient glycemic management has grown. This updated systematic review addressed 10 important aspects of inpatient care for noncritically ill patients with diabetes or newly recognized hyperglycemia.

The increasing numbers of patients who use CSII and CGM in the outpatient setting has led to questions regarding the use of these technologies in the inpatient setting. The systematic review evaluating CGMs provided evidence that compared with point-of-care blood glucose monitoring, implementation of real-time CGM increases the number of patients identified as experiencing hypoglycemia events and decreases mean daily BG values. Supraphysiologic doses of glucocorticoids, which are frequently used in the hospital setting, cause hyperglycemia in the majority of patients, and contribute to adverse outcomes (115, 116). Glucocorticoid-associated hyperglycemia in the hospital requires proactive management with scheduled insulin therapy rather than with “sliding scale” insulin used alone. A variety of scheduled insulin regimens have demonstrated variable efficacy in managing glucocorticoid-mediated hyperglycemia. Many regimens administer NPH at the same time as intermediate-acting glucocorticoids, such as prednisone or methylprednisolone, due to similarities in the pharmacokinetics properties of NPH and these glucocorticoids. The systematic review demonstrated lower mean BG when NPH was added to a basal-bolus insulin regimen, but no difference in hypoglycemia or hospital length of stay. These results should be interpreted cautiously as there was variability in the type, dose, and dosing interval of glucocorticoid therapy, as well as in the insulin regimens used in each of the studies. The optimal insulin regimen for an individual patient will depend on a patient’s nutritional status, as well as the duration, dosing frequency, and choice of glucocorticoid used.

Two small nonrandomized studies evaluating CSII use in the hospital demonstrated that continuation of this therapy for patients using these devices prior to admission may be a reasonable consideration. There is an emphasis on the need for ongoing assessment of a patient’s physical and mental capacity to be able to safely continue to use CSII. There is additional emphasis on having personnel available who are knowledgeable in these technologies who can assess a patient’s ability to continued CSII use on a daily or more frequent basis.

The importance and benefit of Diabetes Self-Management Education and Support (DSMES) is well established, and the time of hospitalization represents an opportunity to begin or reinforce DSMES. This systematic review found that providing survival skill diabetes education in the inpatient setting as part of a comprehensive diabetes discharge-planning program reduces HbA1c at 3 and 6 months and reduces hospital readmissions. There were no differences in patient satisfaction or hospital LOS. A limitation of the existing studies is the pairing of diabetes education with other interventions aimed at improving the transition of the patients from the inpatient to outpatient setting.

Several earlier guidelines recommend various preoperative HbA1c targets for minimizing the risk of postoperative complications in patients undergoing elective surgery (117-119). This systematic review suggests that identified HbA1c measures of < 7% to 8% may be considered a desirable goal for reducing risk for postoperative complications. Hyperglycemia is common in hospitalized patients with and without known diabetes who receive enteral nutrition. Practitioners vary in their preference for insulin regimens that consist of basal insulin, basal-bolus insulin, or NPH insulin, often administered in combination with correctional insulin for glycemic management during enteral nutrition. Consistent with findings from other systematic reviews, the present review found no clinically meaningful differences in daily mean BG, hypoglycemia, or hospital LOS between basal or basal-bolus insulin and NPH-based regimens with accompanying correctional (sliding scale) insulin. This evidence is limited by the observational nature of the available studies, heterogeneity in patient populations, nonrobust study outcomes, and differences in the timing, frequency, and formulations of enteral nutrition.

Insulin therapy is identified as the preferred approach for managing persistent hyperglycemia in hospitalized patients. Due to the lack of efficacy and safety studies in the inpatient setting, oral and noninsulin injectable antidiabetic agents are often discontinued at the time of hospital admission as was recommended in the 2012 Endocrine Society Clinical Guideline. This systematic review compared the effects of insulin and noninsulin therapies on hyperglycemia, hypoglycemia, hospital LOS, and adverse events in hospitalized patients. No RCTs compared insulin therapy to metformin, sulfonylureas, thiazolidinediones, or sodium-glucose cotransporter-2 inhibitors. The 5 RCTs included in this review compared the effects of noninsulin agents (DPP-4 inhibitor, GLP-1 receptor agonists [GLP-1RAs]) and insulin regimens. There was a reduction in the risk of hypoglycemia with both GLP-1RAs and DPP-4 inhibitors. There were no differences in mean daily BG or in cardiac or renal complications between these agents compared with insulin. There was a significant increase in gastrointestinal adverse effects with GLP-1RAs. These findings may not be generalizable due to the selective inclusion criteria in these studies. Furthermore, in the majority of studies of GLP-1RAs and DPP-4 inhibitors, correction insulin was administered for intermittent hyperglycemia. Collectively, these studies suggest that in most adult patients with hyperglycemia hospitalized for a noncritical illness, scheduled insulin therapy remains the preferred therapy for glycemic management. DPP-4 inhibitors used alone or in combination with insulin may be appropriate in select patients with mild hyperglycemia and type 2 diabetes.

Potentially complicating the risk for preoperative hyperglycemia is the practice of providing oral carbohydrate-containing beverages to patients in the immediate preoperative time period. With the exception of a small increase in measures of patient satisfaction reported in 1 study, the systematic review demonstrated no benefit with use of preoperative carbohydrate-containing beverages for patients with diabetes. Nutrition therapy is an important but often overlooked component of inpatient glycemic management. For patients receiving scheduled insulin therapy, adjusting the bolus or prandial insulin component based on the planned carbohydrate content of a meal is referred to as carbohydrate counting (CC). CC is frequently performed by many insulin-treated patients in the outpatient setting, as this permits flexibility in food choices and insulin dosing. There is limited research or published clinical evidence supporting the use of CC in the inpatient setting. The current systematic review did not demonstrate either superiority or inferiority in glycemic measures with CC compared with other methods for prandial insulin dosing in the hospital setting. Two studies reported an increased rate of hypoglycemia with CC, which is difficult to attribute to CC as the events occurred either in the morning or prior to meals. There were no significant differences in hypoglycemia, mean daily BG, or patient satisfaction between the methods of insulin dosing for meals. Lastly, the systematic review addressed the long-debated topic of inpatient use of sliding scale insulin (SSI), defined here as correctional insulin, in the hospital setting. Compared with scheduled insulin therapy, correctional insulin used alone resulted in higher mean daily BG, more frequent episodes of hyperglycemia, a lower risk of hypoglycemia, and no difference in hospital LOS. There was variability among the included studies in study populations and the types of correctional insulin scales used. Despite the lower risk of hypoglycemia with correctional insulin, the risk of uncontrolled severe hyperglycemia (BG values > 300 and > 400 mg/dL) was higher, and most pronounced in patients using insulin therapy prior to admission. These results are in general consistent with other systematic reviews (120-122).

Strengths and Limitations

The strengths of this systematic review relate to the strict adherence to the prespecified protocol and the comprehensive literature search. Other strengths of the study include selecting and appraising studies by a pair of independent reviewers and collaborating with experts from the Endocrinology Society.

A major limitation of the evidence base is the nonrandomized nature of many of the available studies. Varying terminology (eg, sliding scale) and changing technology is another. For example, the inpatient use of CGM and insulin pump therapy were examined independently of each other. There is a growing number of patients using hybrid closed-loop insulin delivery devices paired with real-time CGM, making this an area of increased interest in the hospital setting and for which there is limited data to support this practice at this time (123).

Conclusion

This systematic review has summarized the best available evidence on the management of hyperglycemia in adult patients hospitalized in noncritical care settings and patients with diabetes undergoing elective surgical procedures. The certainty of evidence supporting many hyperglycemia management decisions is low, emphasizing the importance of shared decision-making and consideration of other decisional factors. A guideline panel from the Endocrine Society will incorporate such factors into development of evidence-based practice recommendations.

Glossary

Abbreviations

BG

blood glucose

CGM

continuous glucose monitoring

CSII

continuous subcutaneous insulin infusion

DPP-4

dipeptidyl peptidase-4

GLP-1

glucagon-like peptide-1

GLP-1RA

glucagon-like peptide-1 receptor agonist

GRADE

Grading of Recommendations, Assessment, Development, and Evaluation

HbA1c

glycated hemoglobin

LOS

length of (hospital) stay

NPH

neutral protamine hagedorn (isophane insulin)

RCT

randomized controlled trial

RR

relative risk

Data Availability

Original data generated and analyzed during this study are included in the repository listed in refulgences. The data that support the findings of this study are openly available in “figshare”. https://figshare.com/s/92ee0dc108e2c39ccbda.