Glucose control after glucocorticoid administration in hospitalized patients – a retrospective analysis

Background Glucocorticoid (GC)-induced hyperglycemia is a frequent adverse effect in hospitalized patients. Guidelines recommend insulin treatment to a target range of 6–10 mmol/L (108–180 mg/dl), but efficacies of particular regimes have not been well-studied. Methods In this retrospective cohort study, hospitalized patients receiving GCs at the medical ward were analyzed by treatment (basal-bolus vs. bolus-only vs. pre-mixed insulin) and compared to a non-insulin-therapy reference group. Coefficients of glucose variation (CV), percentage of glucose readings in range (4–10 mmol/L (72–180 mg/dl)) and hypoglycemia (< 4 mmol/L (< 72 mg/dl)) were evaluated. Results Of 2424 hospitalized patients receiving systemic GCs, 875 (36%) developed GC-induced hyperglycemia. 427 patients (17%) had a previous diagnosis of diabetes. Adjusted relative risk ratios (RRR) for the top tertile of CV (> 29%) were 1.47 (95% Cl 1.01–2.15) for bolus-only insulin, 4.77 (95% CI 2.67–8.51) for basal-bolus insulin, and 4.98 (95% CI 2.02–12.31) for premixed insulin, respectively. Adjusted RRR for percentages of glucose readings in range were 0.98 (95% Cl 0.97–0.99) for basal-bolus insulin, 0.99 (95% Cl 0.98–1.00) for premixed insulin, and 1.01 (95% Cl 1.00–1.01) for bolus-only insulin, respectively. Adjusted RRR for hypoglycemia was 13.17 (95% Cl 4.35–39.90) for basal-bolus insulin, 8.92 (95% Cl 2.60–30.63) for premixed insulin, and 2.99 (95% Cl 1.01–8.87) for bolus-only insulin, respectively. Conclusions Current guidelines recommend a basal-bolus regimen for treatment of GC-induced hyperglycemia, but we found similar outcomes with pre-mixed and bolus-only insulin regimens. As GC-induced hyperglycemia is a frequent issue in hospitalized patients, it might be reasonable to prospectively study the ideal regimen. Supplementary Information The online version contains supplementary material available at 10.1186/s12902-021-00914-3.


Background
Glucocorticoids (GCs) are used therapeutically to treat various diseases due to their immunosuppressive effect [1]. Despite their benefits, adverse effects are common. GC-induced hyperglycemia has an approximate incidence of 30% [2]. It is caused by several mechanisms including insulin resistance, increased gluconeogenesis and beta-cell dysfunction [3]. In-hospital hyperglycemia is generally associated with several adverse events such as higher in-hospital mortality, increased length of stay, and transfer to intensive care units [4]. Therefore, optimized diabetes therapy is crucial.
The current guidelines for the treatment of in-hospital hyperglycemia also cover the management of GCinduced hyperglycaemia. As such, the same target range and management is proposed as for all other causes of in-hospital hyperglycemia [5]. Over the last 10 years, the guidelines have changed from a rather tight control of blood glucose levels to modified regimens with at the lower ranges.
This change of paradigm is based on studies showing that a tight control, albeit leading to fewer infections, has no benefits in terms of mortality, myocardial infarction or stroke [5]. Adversely, stricter control entails an increased risk of hypoglycemia [5][6][7]. For persistent GCinduced hyperglycemia, basal-bolus insulin therapy is recommended [7,8].
Hence, we retrospectively investigated different treatment strategies in hospitalized patients with GC-induced hyperglycemia. We were especially interested whether glucose control and hypoglycemia differed depending on the treatment strategy.

Setting and patients
This was a retrospective observational study conducted at the Medical University Clinic in Aarau, Switzerland. The retrospective analysis was done using electronic health record (EHR) data from hospitalized patients treated with GCs between January 2014 and April 2018. The Ethics Committee (EKNZ, Ethics Committee of Northwestern and Central Switzerland) approved the study (No. 2018-01271). The study adheres to the principles of the Declaration of Helsinki and to the STROBE statement. Written informed consent from each patient for the use of EHR data for research was granted by January 1st 2018, as required by local legislation. The Ethics Committee approved use of retrospective data for this analysis in the time frame prior to this date.
Inclusion criteria were age of at least 18 years, a hospital stay of at least 3 days, stay at a medical ward, administration of at least 10 mg prednisolone (or its equivalent), and for hospitalizations after January 1st 2018 general consent for the use of EHR. The sole exclusion criterion was a recurrent admission within 30 days of the index hospitalization. GC dose, indication for GC administration, glucose measurements, insulin dosage, previous diagnosis and treatment of diabetes, and hospital events were evaluated. Interviews conducted 30 days after discharge for quality control reasons were available for evaluation.
Standard of care at the studied hospital At our center, the standard of care after GC administration included blood glucose measurements at 10:00 a.m. and at 2:30 p.m. in order to screen for GC-induced hyperglycemia. This includes a standardized electronic insulin protocol with a low threshold of prescribing bolus insulin, in order to cover meals with an assumed additional need of insulin during GC administration, as published elsewhere [9] and daily screening for hyperand hypoglycemia by endocrinology service is done on weekdays. We also used a low threshold of prescribing bolus insulin, in order to cover meals with an assumed additional need of insulin during GC administration [9].
If GC-induced hyperglycemia occurred, the subsequent treatment choice was made individually, based on hypoglycemia risk, comorbidities, and patient resources to administer the treatment regimen after discharge.
As a local standard, dosage of bolus insulin was adjusted to carbohydrate intake. Basal-bolus insulin usually consisted of insulin detemir twice per day and a rapid acting insulin analogue (insulin lispro or insulin aspart). Typically, the amount of basal and bolus insulin was balanced evenly. In case of pre-mixed insulin treatment, a 50:50 mixture of insulin lispro protamine suspension and insulin lispro injection (HUMALOG® Mix 50/50™ or similar) was used. Two thirds of the total daily premixed insulin dose was generally administred before breakfast, while the remainder, one-third, was injected before lunch.

Outcomes and endpoints
The objective of this study was to evaluate the differences in glucose control between different treatment strategies for GC-induced hyperglycemia. Primary endpoints were glycemic variability, measured by coefficient of variation (CV), percentage of glucose values in the range of 4-10 mmol/L (72-180 mg/dl), and hypoglycemia below 4.0 mmol/L (72 mg/dl). Secondary endpoints were cumulative GC-dose, duration of GC administration, indication for GC administration, mean glucose values and mean insulin administration per patient according to body weight and day.

Hypothesis
We hypothesized that due to individualized treatment, the differences in blood glucose control between the different treatment regimens would be small, but that hypoglycemia rate and CV would be higher in more intensly treated groups.

Statistical analysis
Discrete variables are expressed as counts (percentages) and continuous variables either as means and standard deviations (SD) or as medians and interquartile ranges (IQR).
Continuous variables were analyzed using ANOVA (Analysis of Variance) or Wilcoxon rank-sum test. Categorical and binary variables were analyzed using Chisquare test.
The following three treatment groups were compared to a no insulin treatment reference group: bolus-only insulin, basal-bolus insulin, and premixed insulin.
We evaluated associations between treatment strategy, blood glucose control, and outcomes using a multinomial (polytomous) logistic regression model taking the no insulin treatment group as reference. First, we calculated an unadjusted model, one for each of the three outcomes glycemic variability (glucose CV), percentage of glucose values in range, and hypoglycemia. The CV for glycemic variability was evenly distributed into tertiles and compared to the first tertile as reference.
In a second model, we adjusted each of the outcomes for length of hospital stay (LOS), glucose levels at hospital entry, age, Charlson Comorbidity Index [10], cumulative GC-dose per kilogram of body weight per GC administration day, GC-induced hyperglycemia, and preexisting diabetes.
Significance level was set to an alpha of 5%. All statistical tests were two-sided. We performed no adjustment for multiple testing. The data was analyzed using Stata v15.1 (StataCorp, Texas, USA).

Definitions
GC-induced hyperglycemia was defined as morning fasting blood glucose of > 7.0 mmol/L (126 mg/dl) or a random blood glucose of > 11.0 mmol/L (198 mg/dl) after the start of GCs. Normoglycemia was defined as fasting blood glucose of 7.0 mmol/L (126 mg/dl) or lower, and random or postprandial blood glucose readings of 11.0 mmol/L (198 mg/dl) or lower. GC-induced diabetes was defined by fulfilling the above criteria without a preexisting diabetes. Day 1 was defined as the first day of GC administration. Glycemic variability was calculated as CV = SD/mean*100. The percentage of glucose values in range was defined as the number of finger stick glucose measurements between 4.0-10.0 mmol/L (72-180 mg/ dl). Any blood glucose < 4 mmol/L (< 72 mg/dl) was defined as hypoglycemia. Severe hypoglycaemia was defined as blood glucose < 2.5 mmol/L (< 45 mg/dl). The GC dose was assessed as mg prednisolone equivalent per kg body weight per GC administration day.

Baseline characteristics
Of the 25,183 hospitalized patients during the studied time range, 16% of patients (n = 4060) received GCs of at least 10 mg prednisolone equivalent. 1636 patients had to be excluded from the study either due to a LOS of less than 3 days (n = 986), no given general consent (n = 109) or no follow-up interview (n = 541). Data from 2424 patients were available for analysis (see Fig. 1 for study flow chart).
In the adjusted model, the RRR for a glucose value in range was 1.01 (95% Cl 1.00-1.01) with bolus-only insulin, 0.98 (95% Cl 0.97-0.99) with basal-bolus insulin, and 0.99 (95% Cl 0.98-1.00) with premixed insulin. The RRRs for percentage of glucose values in range in both GC-induced diabetes and preexisting diabetes were significantly higher with all three insulin treatments. (for detailed values, see Fig. 2B and supplemental Table 1).
In the adjusted model, the RRR was 2.99 (95% Cl 1.01-8.87) with bolus-only insulin, 13.17 (95% Cl 4.35-39.90) with basal-bolus insulin, and 8.92 (95% Cl 2.60-30.63) with premixed insulin. The RRRs for GC induced diabetes and preexisting diabetes were significantly higher with all three insulin treatments. (see Fig. 2C and supplemental Table 1 for detailed values). Hypoglycemia was more frequent with higher CV of glucose (see supplemental Fig. 2).

Discussion
In this retrospective observational study, we found that patients with higher glycemic variability tended to have lower glucose values in target range and had a higher relative risk of hypoglycemic events. These two characteristics were more prominent in patients who received basal-bolus or premixed insulin than in patients who received bolus insulin with meals or for correction only. When comparing basal-bolus insulin with premixed insulin, glucose variability was similar, but the premixed insulin group had an 8% higher percentage of glucose values in the target range, alongside a trend towards a lower risk of hypoglycemia (19.1% vs. 13.3%). Although mean insulin per patient was 0.12 E/kg body weight, patients in the bolus-only group needed just 0.05 E/kg of body weight, whereas patients in the basal-bolus group and the premixed insulin group needed 0.23 E/kg body weight and 0.24 E/kg body weight, respectively. As a limitation, we cannot exclude that this allegedly favorable effect of premixed insulin was based on a preselection of suitable patients. Guidelines for hospital hyperglycemia recommend basal-bolus insulin for hyperglycemia but state that for short duration of GCs prandial insulin or intermediate-acting insulin may be sufficient, while more severe hyperglycemia may need to be treated with more complex insulin regimens [7,8]. Several randomized controlled trials (RCTs) [11][12][13] and a recent systematic review [14] advocate premixed or intermediate-acting insulin. The rationale for premixed or intermediate-acting insulin is based on the pharmacologic effect of prednisone on glucose metabolism, the most frequently used GC. Prednisone is generally expected to have a 16-h effect, with a peak after 4-6 h, with low morning glucose, high postprandial values due to glucose intolerance, and a decreasing effect of insulin resistance after 16 h, i.e. in the evening, with a predisposition to evening or nocturnal hypoglycemia [14]. Lakhani et al. showed that with an additional correction accounting for dose, type of GC, and patient body weight, better glucose control is possible [15]. Two RCTs investigated the additional use of oral antidiabetics in hospitalized patients with GC-induced hyperglycemia [16,17]. Long-term data of metformin for the prevention of metabolic complications of GCs have shown some benefits [18,19], but feasibility in hospitalized patients has only been investigated in a small trial [17]. For mild GC-induced hyperglycemia, metformin or other oral antidiabetics are promising, although further studies are warranted to show not only safety, but also a beneficial short-term effect. Our own data were too heterogenous to assess the effect of oral antidiabetics on glucose control.

Limitations and strengths
Our work has several limitations. These result primarily from the fact that the data were from a single center with similar providers and standard operating procedures, thus these standard operating procedures without randomized treatment allocation may have led to confounding by indication. There is also a risk of residual bias from unmeasured factors. Therefore, our results should be interpreted with caution and should serve as the basis for designing a thorough randomized controlled trial. In addition, the relatively small number of patients diagnosed with GC-induced diabetes prohibits a separate analysis of this subgroup. In addition, the fact that blood glucose levels were measured only twice per day is certainly a limitation. It is possible that hypoglycemia could almost have been noticed earlier and could have been treated or prevented earlier. The relatively small number of patients also resulted in a wide confidence interval. Therefore, the results must be interpreted with caution when comparing the groups.
One strength of this study lies in the relatively large number of complete data sets for analysis and that it bases on real-life data, despite its retrospective character. Another strength is that it is one of the first comprehensive reviews investigating associations between treatment strategy, blood glucose control and outcomes.

Conclusions
We found two phenotypes of GC-induced hyperglycemia: a milder type which may be treated with no or low doses of insulin, and a severe type which requires basal-bolus or premixed insulin. In the severe type, basal-bolus and premixed insulin were similarly efficient for achieving treatment targets. Current guidelines recommend a basal-bolus regimen for treatment of GCinduced hyperglycemia. As GC-induced hyperglycemia is a frequent issue in hospitalized patients, it might be reasonable to prospectively study the ideal regimen. We are aware that the results of our retrospective work can be used to influence practice. However, they can guide planning of a sufficiently powered randomized controlled trial which in turn provides reassuring evidence to influence guidelines for patients GC-induced hyperglycemia.
Additional file 1: Supplemental Table 1. Results of the unadjusted and adjusted multinomial logistic regression. Supplemental Fig. 1. Glucose variability according to treatment strategy. Supplemental Fig. 2. Glucose variability in patients with and without hypoglycemia.