Wearable Diabetes Technology for Hospitalized People With Diabetes and End-Stage Kidney Disease, Peripartum State, and Steroid Use

Abstract

Inpatient hyperglycemia remains a challenge, as conventional insulin regimens often lead to both hyperglycemia and hypoglycemia. Traditional glucose monitoring methods, such as point-of-care testing, fail to detect diurnal and nocturnal glycemic fluctuations, contributing to suboptimal control. This review examines the effectiveness of continuous glucose monitoring (CGM) and automated insulin delivery (AID) systems in managing diabetes in hospitalized patients, including those with additional challenges such as end-stage kidney disease (ESKD), pregnancy, and steroid use. In patients with ESKD, CGM has demonstrated reliable glucose measurements and improved glycemic control, particularly in those undergoing hemodialysis. It has been shown to increase time in range (TIR) and reduce hypoglycemia, with clinical accuracy verified in multiple studies. Existing evidence shows that AID systems may offer improved outcomes in this population, with increased TIR and reduced glycemic variability compared with conventional insulin therapy. Continuous glucose monitor use has been beneficial for maternal glycemic control in pregnancy, leading to lower HbA1c levels, increased TIR, reduced maternal hypoglycemia, reduced neonatal hypoglycemia, and admissions to intensive care. Limited studies have evaluated AID system use during labor. In addition, CGM helps identify postprandial hyperglycemia in patients with glucocorticoid-induced hyperglycemia, which is crucial for managing glucose fluctuations. Studies in patients receiving glucocorticoids have shown that continuous glucose monitoring improves glycemic control without significantly increasing hypoglycemic events. In conclusion, limited studies have shown the role of CGM and AID systems and their effects on glycemic outcomes in hospitalized patients with diabetes, particularly those with ESKD, in pregnancy, and those receiving glucocorticoids. These technologies used for glucose monitoring and insulin delivery could offer an alternative method of diabetes management in certain inpatient populations.

Introduction

There has been an increase in hospitalizations of people with diabetes over the last two decades. ¹ Despite advances, hyperglycemia continues to be common in critical and noncritical settings.^2,3 Inpatient hyperglycemia is associated with worse outcomes, including longer hospitalization, increased rates of intensive care unit admissions, higher readmission rates after discharge, and increased mortality. ⁴ The use of conventional insulin therapy to manage hyperglycemia increases the risk of hypoglycemia; which, similar to hyperglycemia, is associated with complications. ⁵ Changes in dietary pattern, nutritional support, glucocorticoid (GC) use, changes in kidney function, or metabolic responses from ongoing illness further compound the challenge of achieving inpatient glycemic targets.

Current inpatient glycemic monitoring relies on multiple point-of-care (POC) glucose testing, ⁶ which fails to provide data between glucose checks, thus missing clinically significant diurnal and nocturnal hyperglycemia and hypoglycemia. The use of diabetes technology in the inpatient setting is an encouraging development. Randomized controlled studies evaluating continuous glucose monitor (CGM) use in the non-critically ill setting have shown safety of use ⁷ and improved identification of glycemic excursions, particularly recurrent hypoglycemic events. ⁸ A recent meta-analysis of available literature showed that use of CGM in noncritically ill patients improves time in range (TIR, 70-180 mg/dL), time below range (TBR, glucose <70 mg/dL), and time in severe hyperglycemia (>250 mg/dL). ⁹ Continuous glucose monitor use in hospitalized people with diabetes was shown to be safe and effective in guiding insulin therapy and is on the rise. ¹⁰ Similarly, pilot studies of automated insulin delivery (AID) use in the hospital have shown significant improvements in glycemic metrics such as TIR, time above range (TAR, glucose >180 mg/dL), mean glucose, and standard deviation of mean glucose (a marker of glucose variability) in the inpatient setting.^11,12 Automated insulin delivery use in the perioperative setting can also reduce the workload of the staff involved in glycemic management. ¹³ However, many such studies on the use of diabetes technology exclude populations that have higher requirements for glycemic management, increased risk for complications, and concomitant comorbidities. In this article, we review the data on current technological advances in inpatient diabetes management in populations having end-stage kidney disease (ESKD), women in labor, and people receiving GCs. We have summarized key studies in Tables 1 and 2.

Table 1.

List of Studies Evaluating Accuracy of Continuous Glucose Monitor in the Inpatient Setting for People With Diabetes and ESKD, Pregnancy, or Glucocorticoid Use.

Study type	Population	CGM	Comparator	MARD (SD)
Retrospective 14	Type 2 diabetes and ESKD	• Freestyle Libre 2 • Dexcom G6	• Accuchek Guide • Nova StatStrip • Freestyle Optium Neo H	12.7% (12.9)
Prospective 15	Type 1 or type 2 diabetes and ESKD	Dexcom G6	Nova StatStrip	20.5% (13.3)
Prospective 16	Type 1 or type 2 or gestational diabetes and undergoing C-section	Freestyle Libre 2	Nova StatStrip	9.3%
Prospective 17	Type 1 or type 2 diabetes and glucocorticoid use	Freestyle Libre	Medisafe Fit	16.5% (5.6)
Prospective 18	Type 2 diabetes or LADA or stress hyperglycemia and glucocorticoid use	Dexcom G6	Accuchek Inform II	13.9%

Abbreviations: CGM, continuous glucose monitor; MARD, mean absolute relative difference; SD, standard deviation; ESKD, end-stage kidney disease; LADA, latent autoimmune diabetes in adults.

Table 2.

List of Studies Using Diabetes Technology Devices and Their Effects in People With Diabetes and ESKD, or Women Undergoing Labor, or Glucocorticoid Use in the Inpatient Setting.

Study type	Population	Intervention	Findings
Randomized, controlled trial (post hoc analysis) 19	Type 2 diabetes and ESKD on hemodialysis	MPC algorithm with Dana R Diabecare	• ↑ TIR by 38% • ↓ TAR by 37% • ↓ Mean Glucose by 52 mg/dL • ↓ Glucose SD by 23 mg/dL • ↓ Coefficient of Variation
Prospective, observational (pooled analysis) 20	Type 1 diabetes and labor	Florence D2W or D2A algorithm with Dana R Diabecare	• No correlation in maternal glucose, Time 63-140 mg/dL, and Time >140 mg/dL with neonatal hypoglycemia • No maternal hypoglycemia during labor or 48 hrs postpartum • Safety of use during 48 hrs postpartum
Randomized, controlled trial 21	Type 1 diabetes and labor	Medtronic Minimed 780G	• ↑ in Time 63-140 mg/dL by 8% • ↑ in TIR by 5% • No difference in neonatal/maternal hypoglycemia or NICU admissions
Randomized, controlled, single-blind, trial 22	Type 1 or type 2 diabetes and glucocorticoid use in renal transplant	Medtronic Guardian 3	• ↓ in mean glucose by 13 mg/dL • ↓ hyperglycemia episodes

Abbreviations: ESKD, end-stage kidney disease; MPC, model predictive control; TIR, time in range (glucose range 70-180 mg/dL); TAR, time above range (glucose range >180 mg/dL); SD, standard deviation; hrs, hours; NICU, neonatal intensive care unit.

End-Stage Kidney Disease

Glycemic control is challenging in people with diabetes and ESKD, and this is further complicated in the hospital setting. A retrospective study showed that more than half of those with type 2 diabetes (T2D) receiving hemodialysis experienced inpatient hypoglycemia, with a majority occurring the day before a hemodialysis session. ²³ This could be related to changes in kidney function, reduced oral intake, co-administered medications, and ongoing illness. In addition, CGM use in people with T2D and ESKD on hemodialysis can reveal previously undetected glycemic variability and lead to a reduction in hypoglycemia. ²⁴ However, such studies were conducted in the outpatient setting or with older CGM models. As there is greater glucose variability in hospitalized patients undergoing hemodialysis, ²³ the applicability of results from outpatient-based studies has limited relevance in the inpatient setting.

To establish the reliability of CGM in hospitalized patients undergoing chronic hemodialysis, a multicenter, retrospective study evaluated people with type 1 diabetes (T1D) using Dexcom G6 (San Diego, California) and Abbott FreeStyle Libre 2 (Alameda, California). ¹⁴ A mean absolute relative difference (MARD) of 12.7% was noted with reference to POC glucose testing. The %15/15, %20/20, and %30/30 agreement rates were 67%, 76%, and 90%, respectively. In consensus error grid analysis, 99.2% of glucose pairs were in zones A/B, indicating good clinical accuracy. As participants were hospitalized, an accurate measurement of POC glucose testing could be conducted between hemodialysis days and the MARD differences between interdialytic and intradialytic days were found to be minimal. Similarly, in a prospective observational study of noncritically ill hospitalized patients with diabetes on maintenance hemodialysis, a strong correlation was noted between CGM (Dexcom G6) and POC glucose using scatter plots. ¹⁵ Although the MARD was 20.5%, the consensus grid analysis showed greater than 98% of CGM values in zones A and B. No differences were noted in either analysis between hemodialysis days and nondialysis days. These studies indicate the utility of CGM with acceptable accuracy in hospitalized patients with diabetes with ESKD.

A retrospective study assessed the relationship between hemoglobin A1c (HbA1c) and the glucose management indicator (GMI) in hospitalized patients with diabetes-related kidney disease using insulin pumps. ²⁵ The authors reported no correlation between HbA1c and GMI in patients with stage 4 and 5 diabetic kidney disease not undergoing dialysis. Similar results have been observed in patients with diabetes and ESKD undergoing hemodialysis in the ambulatory setting. ²⁶ The study also showed that patients with advanced diabetic kidney disease tended to have more TBR as compared to those with less severe disease. Thus, CGM-derived glycemic measurements, such as mean glucose or TIR, should be preferentially used in this patient population, as CGM measurements of interstitial glucose are as reliable as POC glucose, and HbA1c is known to underestimate glycemic control in those with ESKD.

Prospective studies in the outpatient setting have shown that CGM is a useful tool for improving glycemic outcomes in people with ESKD due to more frequent therapeutic changes. These studies, in both people with T1D and T2D, demonstrated improved glycemic control with CGM use.^27,28

Data on using an automated insulin delivery (AID) system in hospitalized patients with ESKD and diabetes mellitus is sparse. However, in a post hoc analysis of a randomized controlled trial (RCT), fully automated closed-loop delivery was compared with conventional insulin therapy in hospitalized patients with T2D undergoing hemodialysis. ¹⁹ Higher TIR was noted for those using an AID system compared with conventional therapy (69% vs 31.5%, P < .001). Similar improvements were seen with mean glucose and TAR in those using an AID system. Although there were no differences between times in hypoglycemia or hypoglycemia events, glycemic variability (as measured by the standard deviation of mean glucose and coefficient of variability) significantly improved in the AID group.

In conclusion, the existing evidence supports the use of CGM as a reliable tool for glycemic monitoring in hospitalized people with diabetes and ESKD, demonstrating acceptable accuracy in clinical settings. In addition, AID systems substantially improve glycemic outcomes and could be continued for AID users with ESKD and diabetes during hospital admission. However, further longitudinal studies are needed to evaluate long-term clinical outcomes and the effectiveness of these technologies in improving glycemic control in hospitalized patients with diabetes and ESKD.

Peripartum State

During pregnancy, a key goal is the avoidance of maternal hyperglycemia and hypoglycemia. Insulin resistance increases around 16 weeks of gestation, and total daily insulin requirements increase through week 36.^29,30 The peripartum period refers to the time shortly before, during, and immediately after childbirth and is generally defined as the time from labor onset, continuing through the immediate postpartum phase, up to 24 to 72 hours after delivery. During the peripartum period, several physiological changes affect glucose metabolism with a marked increase in glucose utilization and production. The increase in glucose during labor is associated with elevated levels of insulin, glucagon, cortisol, and catecholamines, which promotes greater glucose production and utilization. ³¹ With the delivery of the placenta, insulin sensitivity increases significantly, leading to a rapid, immediate reduction in insulin requirements postpartum. This increased insulin sensitivity gradually returns to pre-pregnancy levels over one to two weeks. ³² Neonatal complications such as large for gestational age, preterm delivery, neonatal hypoglycemia, and perinatal death are common in infants born to mothers with T1D. ³³ The use of GCs prior to planned delivery for fetal lung maturation also affects maternal glycemic control. Such dynamic changes pose additional challenges to optimal glycemic control in the peripartum period. ³⁴

With advancements in reliability and accuracy of CGM, its use has been increasing over the last decade amongst pregnant women with pre-existing or gestational diabetes. ³⁵ An observational, prospective study evaluated CGM accuracy in pregnant women with gestational or pregestational insulin-treated diabetes undergoing cesarean deliveries. ¹⁶ An intermittently scanned CGM (Freestyle Libre 2) showed a MARD of 9.3% and 100% of the values in zones A and B of the Clarke Error Grid analysis, confirming good intrapartum accuracy consistent with data from nonpregnant individuals.

Outpatient CGM use and its effects on maternal and neonatal outcomes have been evaluated by various studies. A RCT showed that pregnant women with T1D using CGM during pregnancy had a significant reduction in HbA1c, an increase in TIR, and a decrease in maternal hypoglycemia compared with those with POC glucose testing. ³⁶ This resulted in better neonatal outcomes, including lower incidence of large for gestational age, fewer neonatal intensive care admissions, and reduced neonatal hypoglycemia, likely due to reduced exposure to maternal hyperglycemia.

A secondary analysis of the same study and another observational prospective study of people with T1D reported intrapartum glycemic metrics.^37,38 In the intrapartum period (defined in this analysis as 24 hours prior to delivery), participants had a mean glucose of 113 to 115 mg/dL, 72% to 82% time-in-target (70-140 mg/dL), 15% to 19% time >140 mg/dL, and 0% to 9% time <70 mg/dL. In pregnant women with gestational or pregestational diabetes, another prospective observational study showed similar intrapartum glycemic profiles. ³⁹ The mean glucose was 102 mg/dL, the coefficient of variation was 13.8%, TIR (70-110 mg/dL) was 62.1%, time above range was 27.5%, and time spent in hypoglycemia was 0%.

Several studies have evaluated the relation between maternal intrapartum glycemic control and neonatal hypoglycemia and found no correlation.^37-40 In addition, CGM use was not associated with reduced rates of neonatal hypoglycemia. ³⁷ There are mixed results on intrapartum glycemic control and preterm delivery which may be due to differences in how intrapartum state is defined across studies.^37,39 One study showed that mothers with an intrapartum time spent in glucose >110 mg/dL of 61% or more had a higher rate of preterm delivery, higher cord C-peptide levels, infant use of airway support, use of insulin infusion during labor, and postpartum hemorrhage. ³⁹

In line with the promising outcomes reported with CGM use during pregnancy, AIDs have been evaluated in pregnant women with diabetes. In pregnant women with T1D, AID systems were associated with better time-in-target range (63-140 mg/dL), mean glucose, time spent in glucose >140 mg/dL, and reduction in hypoglycemia compared with standard therapy.^41,42 These studies included few participants; however, the prospective observational design confirms the safe use of AIDs in pregnancy. Data on intrapartum use of AIDs are available from secondary analyses of prospective observational trials. A pooled secondary analysis of 2 prospective observational trials showed that intrapartum AID users had a median time-in-target range of 82%, time >140 mg/dL of 16%, no time in hypoglycemia, and a mean glucose of 124 mg/dL. ²⁰ No correlation between neonatal hypoglycemia and intrapartum glycemic metrics (mean glucose, time-in-target range, or time-above-target) was found. When comparing intrapartum AID use with standard insulin therapy in another secondary analysis, AID users (who used AID during pregnancy and continued AID during labor) had higher time-in-target range (71% vs 63.1%, P = .03) with no differences in time spent in glucose >140 mg/dL, time in hypoglycemia, or mean glucose. Despite higher time-in-target range for the AID group, there were no differences in the rates of neonatal hypoglycemia or neonatal intensive care unit admissions. ²¹

Given the reassuring data on CGM and AID use, recent consensus guidelines from professional societies recommend that health care professionals advise pregnant women to continue using a CGM during hospitalization to identify glucose trends and prevent hypoglycemia or hyperglycemia. ³⁴ These guidelines also underscore the need for further studies to assess the accuracy of CGMs during pregnancy and the peripartum period. ⁴³ Hospital policies could adapt to allow parturient women to continue pre-existing CGM use during the intrapartum period. This experience can inform the development of clinical workflows for broader inpatient CGM use and enhance data collection. The use of AID systems during labor and delivery should be individualized. Glucose management during labor and delivery can be challenging for some patients, especially if their care teams have limited experience with diabetes technology and when GCs are used in preterm delivery for fetal maturation.

GC Use

Glucocorticoids are commonly used in hospitals for their potent anti-inflammatory and immunosuppressive effects. Conditions such as chronic obstructive pulmonary disease (COPD) exacerbations, gout flares, allergic reactions, autoimmune diseases, organ transplantation, certain chemotherapeutic regimens, and COVID-19 infections utilize supraphysiologic doses of GCs as part of the typical treatment plan. A common unintended consequence is steroid-induced hyperglycemia, occurring in 20% to 70% of patients without diabetes and >50% of those with diabetes.^44-47 Glycemic control in this patient population is crucial, as severe hyperglycemia in the inpatient setting has been associated with an increased risk of hospital complications, infections, longer hospital stays, and mortality.^48,49

Steroid-induced glucose elevations occur via GC receptor activation across multiple tissues. Glucocorticoids inhibit insulin production and secretion from pancreatic beta cells, ⁵⁰ induce insulin resistance within muscles, leading to decreased glucose uptake, ⁵¹ while in hepatocytes, GCs promote key enzymes involved in gluconeogenesis. ⁵¹ Glucocorticoids have been posited to potentiate the effects of other glucose-stimulating hormones, such as glucagon and epinephrine. ⁵² Last, in adipocytes, steroids increase lipolysis and the release of free fatty acids (FFAs), further exacerbating insulin resistance in other tissues, particularly muscle and liver. ⁵³

The effect of a specific steroid regimen on glycemic control can be challenging to predict and depends greatly on multiple factors including the type of steroid, dosage, administration route, duration of treatment, nutritional intake, and the individual’s baseline glucoregulatory capacity (eg, healthy, prediabetes, diabetes).^46,54 Short- or intermediate-acting systemic GCs, such as hydrocortisone or prednisone, taken only once daily in the morning, primarily induce post-prandial hyperglycemia, with glycemia levels rising throughout the afternoon and evening. However, these GCs typically have minimal impact on fasting glucose levels.^54-56 Conversely, supraphysiologic doses of long-acting GCs such as dexamethasone, or multiple daily administrations of shorter-acting GCs, can significantly increase both fasting and postprandial glucose concentrations. ⁵⁷

Continuous glucose monitors have been used to help identify glycemic patterns and improve management of steroid-induced hyperglycemia for both patients with and without diabetes. A CGM study evaluating normoglycemic women receiving large doses of GC as part of gynecologic chemotherapy found that dexamethasone 30 mg daily (prednisone-equivalent dose [PED] 200 mg) resulted in significant hyperglycemia, primarily postprandially lasting into the late evening, with more than 40% of women spending >4 hours with a BG >200 mg/dL. However, a lower dose of 8 mg daily (PED 40 mg) had minimal glycemic elevations, and on the day following the cessation of dexamethasone, the glucose levels returned to normal. Interestingly, the 1 patient with prediabetes (HbA1c 6.2%) demonstrated the most severe hyperglycemia in the cohort, with glycemic levels persisting >200mg/dL on the day following dexamethasone cessation. ⁵⁷

Another study evaluating patients without diabetes receiving methylprednisolone 40 mg once daily (PED 50 mg) in the morning for COVID-19 infection in the hospital found that fasting glucose levels were in the normal range, with levels only starting to rise after lunch consumption and peaking about 1 to 2 hours after dinner (median 175 mg/dL). However, the glycemic distribution was wide among the patients, with a third having a TAR ≥25%, and 21% of patients above >250 mg/dL more than 10% of the time. Conversely, over half the cohort had >90% TIR without receiving any diabetes medications. ⁵⁸

Among patients with diabetes, those receiving a course of once-daily prednisolone (PED 26 ± 9 mg/dL) for COPD flares were found to have higher fasting and postprandial glycemic concentrations as compared to similar patients without diabetes. ⁵⁶ Receiving a single intermittent dose of dexamethasone (eg, a single dose of intravenous dexamethasone every two weeks) can also result in a steady but profound glycemic increase within 3 hours of GC administration and last 24 to 36 hours, before returning to baseline. ⁵⁹

There has been a lack of RCTs using in-hospital CGM to help manage steroid-induced hyperglycemia. A few studies used CGMs to compare the efficacy of different insulin regimens. In one study, hospitalized patients receiving once-daily prednisolone ≥20 mg/day were randomized to NPH vs glargine as their basal insulin. CGM data demonstrated no differences between groups in mean blood glucose, insulin dose, or time outside of target range. ⁶⁰ Similar results were seen in another RCT, evaluating patients with diabetes on multiple daily intermediate-acting GC, with no differences in mean blood glucose, glucose variability, or hospital length of stay (LOS) between NPH and glargine groups. ⁶¹ Lastly, a small crossover study evaluated patients with known diabetes or previous steroid-induced hyperglycemia receiving GC-containing chemotherapy (PED ≥12.5mg for 3-10 consecutive days in each cycle). Patients were randomized to correctional insulin per protocol versus intermediate-acting insulin (IMI) as an add-on to their routine diabetes medication regimen. The IMI treatment had a higher proportion of TIR (34% vs 21%), with no difference in rates of hypoglycemia. ⁶²

To our knowledge, only one prospective RCT has been performed comparing CGM versus conventional POC glucose measurements for steroid-induced hyperglycemia in the inpatient setting. Post-renal transplant patients with diabetes were randomized to one of the two glucose monitoring strategies during the first five days post-transplant. The investigators found that the CGM group had significantly lower median daily glucose levels and fewer episodes of hyperglycemia, with no differences in rates of hypoglycemia, postoperative infections, or LOS. ²²

Data on AID systems use in inpatient GC use come from a prospective observational study where an adaptive proportional derivative–based algorithm (APD) was compared with a fading memory proportional derivative–based (FMPD) algorithm. Adaptive proportional derivative–based algorithm responds to changes in insulin sensitivity, which occur with GC use. The authors showed better glycemic control with an improvement in postprandial glucose increments without an increase in hypoglycemia in APD users. ⁶³

The current evidence indicates the safety and utility of CGM use in the inpatient setting and GC use. Postprandial hyperglycemia is often missed with POC glucose testing following hospital protocols, and CGM use could help identify these hyperglycemia episodes. More studies are needed to evaluate clinical outcomes with CGM and AID use in the inpatient setting for people receiving GC; however, current data suggest the safety of CGM technology.

Conclusion

Technological developments are changing the paradigm of the field of medicine and diabetes care with vast evidence-based results supporting the use of CGM and AIDs in people with diabetes in ambulatory and hospitalized settings. However, there are areas of uncertainty where scientific evidence is lacking. We anticipate and recommend that future studies focus on expanding the use of CGM and AID systems in complex inpatient scenarios where diabetes management is challenging. In ESKD, HbA1c is unreliable, and guidelines now recommend CGM use; however, prospective trials of CGM-guided therapy or closed-loop insulin delivery during dialysis are needed. Pregnancy and the peripartum period present with rapidly changing insulin requirements. More extensive randomized studies involving women with pre-pregnancy type 2 or gestational diabetes are required to develop algorithms for use during labor and delivery. Recent CGM studies reveal marked steroid-induced hyperglycemia, but there is a need for randomized studies for CGM use in the hospital and AID systems that consider steroid use. Increased data availability from studies will support the broader use of diabetes technology in inpatient settings and can help drive necessary changes in hospital policies, where current regulations may be prohibitive. In addition, integrating CGM and AID data into electronic health records will ensure the data is easily accessible for use. Identifying patient populations that would benefit most from inpatient CGM and AID systems and existing institutional barriers are crucial for targeted implementation. No CGM or AID is approved by a regulatory authority worldwide for inpatient or dialysis use. By filling these gaps, CGM and AID can be better tailored to improve glycemic management in patients with ESKD, diabetes in pregnancy, or steroid-associated hyperglycemia.