Hypoglycemia Incidence and Factors Associated in a Cohort of Patients With Type 2 Diabetes Hospitalized in General Ward Treated With Basal Bolus Insulin Regimen Assessed by Continuous Glucose Monitoring

Abstract

Introduction:

Continuous glucose monitoring (CGM) is a better tool to detect hyper and hypoglycemia than capillary point of care in insulin-treated patients during hospitalization. We evaluated the incidence of hypoglycemia in patients with type 2 diabetes (T2D) treated with basal bolus insulin regimen using CGM and factors associated with hypoglycemia.

Methods:

Post hoc analysis of a prospective cohort study. Hypoglycemia was documented in terms of incidence rate and percentage of time <54 mg/dL (3.0 mmol/L) and <70 mg/dL (3.9 mmol/L). Factors evaluated included glycemic variability analyzed during the first 6 days of basal bolus therapy.

Results:

A total of 34 hospitalized patients with T2D in general ward were included, with admission A1c of 9.26 ± 2.62% (76.8 ± 13 mmol/mol) and mean blood glucose of 254 ± 153 mg/dL. There were two events of hypoglycemia below 54 mg/dL (3.0 mmol/L) and 11 events below 70 mg/dL (3.9 mmol/L) with an incidence of hypoglycemic events of 0.059 and 0.323 per patient, respectively. From second to fifth day of treatment the percentage of time in range (140-180 mg/dL, 7.8-10.0 mmol/L) increased from 72.1% to 89.4%. Factors related to hypoglycemic events <70 mg/dL (3.9 mmol/L) were admission mean glucose (IRR 0.86, 95% CI 0.79, 0.95, P < .01), glycemic variability measured as CV (IRR 3.12, 95% CI 1.33, 7.61, P < .01) and SD, and duration of stay.

Conclusions:

Basal bolus insulin regimen is effective and the overall incidence of hypoglycemia detected by CGM is low in hospitalized patients with T2D. Increased glycemic variability as well as the decrease in mean glucose were associated with events <70 mg/dL (3.9 mmol/L).

Hyperglycemia in the hospital setting affects 38-46% of non-critically ill patients and is associated to a higher rate of cardiovascular and infectious complications, increased hospital stay, costs and mortality.^1-5 Randomized control trial has demonstrated that basal bolus insulin regimen achieves optimal glycemic control in non–critically ill patients with T2D.⁶ Nevertheless, the main concern of this therapy is the risk of hypoglycemia, a serious adverse event that leads to cardiovascular and neurologic complications, ICU admission and mortality.^7,8 Bedside capillary POC monitoring is the standard of care to assess glycemic control in the hospital. Diabetes guidelines recommend bedside capillary POC testing before meals and at bedtime to assess glycemic control and to adjust insulin therapy in the hospital.⁹ However, it has been shown to frequently misses detection of hypoglycemia events and to have limited accuracy with a variation up to 20% from blood glucose level.^10,11

CGM measures interstitial glucose every 5-15 minutes, thus providing a more complete glycemic profile during 24-hours compared to standard POC glucose testing.¹² Studies in T2D hospitalized patients treated with basal bolus insulin regimen have shown that CGM improves the detection of diurnal and nocturnal hypoglycemic events compared to POC glucose testing.^11-17 More than 50% of the inpatient hypoglycemic events occurred between dinner and breakfast; suggesting that many of these episodes would be missed by standard POC testing. Schaupp et al reported that the number of nocturnal hypoglycemic episodes <70 mg/dl (<3.9 mmol/L) was 15-fold higher, and the number of episodes >250 mg/dl (>13.9 mmol/L) detected by CGM during night was 12.5-fold higher compared to capillary POC glucose testing in general medicine patients with T2D treated with a basal bolus insulin regimen for ≥3 days.¹⁴ In addition, our group previously reported that CGM detected a higher number of episodes of nocturnal and asymptomatic hypoglycemia compared to POC testing.¹⁸

The aim of the study is to assess efficacy, incidence of alert of hypoglycemia <70 mg/dL (3.9 mmol/L), clinically significant hypoglycemia <54 mg/dL (3.0 mmol/L), and to determine clinical factors associated with hypoglycemic events in patients with T2D treated with basal bolus insulin regimen in general ward. We also describe efficacy, in terms of percentage of time in range, and glycemic variability by using CGM.

Methods

This is a post hoc analysis of a prospective cohort study in insulin-treated adult patients with T2D treated at the Hospital Universitario San Ignacio in Bogotá, Colombia.¹⁸ We excluded surgical patients, and those treated with systemic steroids, enteral or parenteral nutrition, patients requiring intensive care management, chronic liver disease or cirrhosis, creatinine clearance <30 ml/min, pregnancy, mental conditions limiting understanding of the study and patients with diabetic ketoacidosis or hyperosmolar nonketotic state. Patients agreed to participate in the study and signed the informed consent approved by the Ethics Committee of Hospital Universitario San Ignacio and Pontificia Universidad Javeriana.

34 general medicine patients were consecutive selected in random days during two years. On admission, preadmission antidiabetic oral treatment was discontinued. All patients were treated with basal bolus therapy during the study and the entire hospitalization, however the data of CGM was available and analyzed only the first 3-6 days of hospital stay. Sensor was installed the second day of hospitalization and one sensor per patient was used. Basal insulin used was insulin glargine, and the rapid acting insulin was insulin glulisine (Lantus and Apidra, Sanofi-Aventis, Bridgewater, NJ). The initial total daily dose (TDD) was calculated at 0.3-0.5 unit per kilogram according to age, admission POC testing and serum creatinine. 50% of the total daily insulin dose was given as insulin glargine and 50% was divided in three doses of insulin glulisine given before meals. Insulin dose adjustment was made according to the POC testing. Glycemic target was set at 100-140 mg/dL (5.5-7.7 mmol/L) of fasting glucose level and 140-180 mg/dL (7.7-10 mmol/L) of preprandial glucose. In patients with a fasting glucose of 140-180 mg/dL (7.7-10 mmol/L), the basal insulin dose was increased by 10%, whereas in those with a fasting glucose >180 mg/dL (10 mmol/L), it was increased by 20%. Medical staff and patients were both blinded to the CGM data. Calibrations were performed as recommended by the manufacturer using capillary blood glucose measures. The first calibration was done 1 to 3 hours after the insertion of the sensor, afterward calibrations were performed before meals. Food intake was registered by the patient. Insulin doses were injected and registered by the nursery staff.

When CGM was removed, data was downloaded using the Software iPro CareLink 3.0 (Medtronic, Minneapolis, MN). Data collected were transferred to a calculation software in MATLAB®. The data of each patient were organized by calendar days (00:00-23:59 hours). Based on these data, we calculated the glycemic variability metrics.¹⁹

Severe hypoglycemia was defined as cognitive impairment requiring external assistance for recovery. If severe hypoglycemia was documented, a bolus of IV dextrose 250 cc 10% was administered immediately to the patient according to the guidelines of the institution. Event of hypoglycemia <70 mg/dL in an alert patient was managed with an oral load of 15 grams of glucose; a bedside capillary POC testing fifteen minutes after the load was performed. This procedure was performed until bedside capillary POC testing rose at ≥70 mg/dL. Sensor failure was defined as consecutive losses of more than 50 samples and were detected on preprocessing of the records. Lower losses were linearly interpolated.²⁰

The following CGM parameters were analyzed according to the current recommendations:^9,21 hypoglycemia in terms of rate of events, percentage of time and area under the curve (AUC) <54 mg/dL (3.0 mmol/L) and <70 mg/dL (3.9 mmol/L). An event of hypoglycemia was considered after at least 15 minutes below the threshold which is the same as 3 consecutive samples below the respective threshold. If the event lasted more than 120 minutes, another event of hypoglycemia was accounted and it was considered as prolonged hypoglycemia. The end of the event was defined by interstitial glucose above the threshold for more than 15 minutes. Percentage of time in range was considered between 140-180 mg/dL (7.8-10.0 mmol/L), this range was determined according to the target of hospitalized patients recommended by ADA 2018. Hyperglycemia was reported in terms of rate of events, percentage of time and AUC >180 mg/dL (10.0 mmol/L) and >250 mg/dL (13.8 mmol/L). Glycemic variability metrics analyzed were standard deviation (SD) and coefficient of variation (CV). The night segment was considered from 12:00 am to 6:00 am.

A descriptive analysis of the information was performed. The evaluation of hypoglycemia events is presented for the total time of the study and per each day. A Poisson regression was used to describe the association between the number of events <70 mg/dl, mean glucose level and glycemic variability (SD and CV%). A threshold value of 36% was used for CV.²² Finally, a longitudinal analysis to evaluate the change during the study time was performed using a random-effects models using the GLS estimator for each one of outcomes: AUC, % of time, number of events of hyperglycemia >180 mg/dL and >250 mg/dL and of hypoglycemia <54 mg/dL and 70 mg/dL. All models were evaluated with a significance level of .05; analysis was performed in the program Stata 15.0

Results

38 patients were initially considered but four patients were excluded due to sensor failure. Baseline demographic and clinical characteristics of 34 included patients are shown in Table 1. Mean age was 66.1 ± 8.6 years with a duration of diabetes of 14.7 ± 8.9 years. Mean A1c was 9.2 ± 2.6%. The main reason of hospitalization was infection in 14 patients (41.2%) followed by coronary heart disease in 10 patients (29.4%). Twenty patients were on treatment with insulin previous to the admittance to the hospital. The average number of days under CGM was 4.3 ± 1 day. 100% of patients completed 3 days of analysis. In all, 22, 14, and 4 patients completed 4, 5, and 6 days with CGM, respectively. Reasons for shorter CGM included early hospitalization discharge.

Table 1.

Baseline Characteristics of Patients Included in the Study.

Baseline characteristics	n = 34
Age, mean (SD)	66.1	(8.6)
Diabetes duration time, mean (SD)	14.7	(8.9)
Comorbidities, n (%)
Cardiovascular disease	11	(28.9)
Kidney disease	14	(36.8)
Anthropometric measures, mean (SD)
Weight	69.6	(14.1)
BMI	26.5	(4.9)
Previous ambulatory treatment, n (%)
Metformine	16	(47.6)
Sulphonylurea	12	(35.2)
DPP4 inhibitor	2	(5.9)
Basal insulin	20	(58,8)
Prandial insulin	17	(50)
Indication of hospitalization, n (%)
Hyperglycemia	7	(20.5)
Coronary heart disease	10	(29.4)
Infection	14	(41.2)
Others	3	(8.8)
Admission laboratories, mean (SD)
Blood glucose, mg/dl	254.6	(153)
A1c, %	9.2	(2.6)
Creatinine, mg/dl	1.2	(0.4)
Hospitalization outcomes.
Duration of hospitalization in days, mean (SD)	14.9	(11)
Death during hospitalization, n (%)	2	(5)

Incidence of Hypoglycemia Detected by CGM: Incidence of Events, Percentage of Time and Area Under the Curve (AUC)

A total of 14.7% and 5.8% of patients had hypoglycemia <70 mg/dL and <54 mg/dL detected by CGM, with a rate of events <70 mg/dL and <54 mg/dL of 0.323 and 0.059 episodes/patient, respectively. Of events, 60% occurred between dinner and 6 am. Eleven events <70 mg/dL were documented in five patients, two of them were previous to the admission to the hospital on insulin therapy. First TDD in hospital was comparable to the dose previous to the admission. There were no episodes of severe hypoglycemia. The events of alert of hypoglycemia and clinically significant hypoglycemia presented in the first days two days of the basal bolus treatment, subjects who presented alert of hypoglycemia were the same which presented clinically significant hypoglycemia.

The incidence of events, percentage of time, and AUC <54 and <70 mg/dl were low during the follow-up. We compared the results of each day evaluated on CGM with the results of the first day on CGM (Table 2).

Table 2.

Incidence of Events <54 and <70 mg/dL per day, Percentage of Time (%t), and AUC.

	Events <54 mg/dL						Events <70 mg/dL
Day	Incidencea	P	%t	P	AUCb	P	Incidencea	P	%t	P	AUCb	P
1	0.0294		0.06		0.007		0.147		0.58		0.059
2	0.0294	1.00	0.09	.68	0.016	.748	0.059	.009	0.26	.019	0.034	.286
3	0.0294	1.00	0.09	.68	0.016	.748	0.059	.009	0.26	.019	0.034	.286
4	0.0275	.56	0.04	.80	0.017	.739	0.055	.020	0.28	.060	0.028	.247
5	0.0190	.79	0.06	.98	0.019	.938	0.060	.062	0.31	.151	0.035	.455
6	0.0180	.87	0.06	.99	0.027	.964	0.025	.127	0.26	.322	0.034	.644

^aEvents per patient.

^bmg/dL/min per day.

Factors Associated With Hypoglycemia

Based on analyzed CGM data, factors associated with events <70 mg/dL during the hospital stay were mean glucose level, glycemic variability (SD and CV), time since diagnosis.

The incidence of hypoglycemia was decreased in 14% for every 10 mg of increase of in the mean of glucose (IRR 0.86, 95% CI 0.79, 0.95, P < .01). Using the threshold of 36% for the CV the incidence rate increased 3 times in the group of patients with high glycemic variability (IRR 3.12, 95% CI 1.33, 7.61, p <0,01). In a similar way, the incidence increased when the SD increased (Table 3).

Table 3.

Factors Associated With Hypoglycemia (Events <70 mg/dL).

Variable	IRR	IC	P
Mean G	0.86	0.79-0.95	.004
CV%	3.12	1.33-7.69	.009
SD	1.81	1.17-2.80	.007
BMI	0.85	0.69-1.05	.14
TDD	0.99	0.94-1.04	.69
Time since diagnosis DM	0.90	0.83-0.98	.01
Hospital stay	0.84	0.69-1.01	.069

Glycemic Variability

Glycemic variability was analyzed during the first 3 days of CGM use. There was a statistically and progressive reduction on glycemic variability after the first day. During the first day, the SD was 35.0 ± 13.9, and reduced on the second by −6.42 (P < .0116) and on the third day by −7.19 (P < .0046). CV% during the first day was 20.4 ± 5.2%, and reduced on the second and third day by −2.9 (P < .004) and −3.2 (P < .002), respectively. Only one patient remained with a CV% >36% on the second and third day of CGM.

Efficacy

Figure 1 shows the reduction of interstitial glycemic average through hospitalization days. Mean glucose level at the admittance was 254.1 ± 153.9 mg/dL, the first day of treatment it was reduced to a mean glucose level of 177.8 ± 40.0 mg/dL, mean difference was −76.3 mg/dL (P = .003). A progressive and statistical reduction was found, reaching a glucose average in the fourth day of hospitalization of 160 ± 7.7 mg/dL (P = .021). There was a significant increase of percentage of time on targets during the follow-up (Table 4). The average of percentage of time on targets during the hospitalization was 77.57% and total mean interstitial glucose was 170.37 ± 35.2 mg/dL.

Figure 1.

Mean interstitial glucose level per day.

Table 4.

Percentage on Time in Range (140-180 mg/dL) During the Follow-Up.

Day	Percentage of time in range	IC 95%	P
1	71.6
2	77.3	0.37, 10.9	.036
3	78.1	0.26, 12.7	.041
4	78.9	0.15, 14.8	.055
5	87.2	3.17, 27.9	.014

Table 5 shows the incidence of events, percentage on time and AUC >180 mg/dL and >250 mg/dL per day. Comparison of each day evaluated is performed with day 1 of CGM. Data of five days on CGM showed an incidence of 8.7 events >180 mg/dL per patient. Patients were exposed to an interstitial glucose value >180 mg/dL in 38.3% of time and AUC >180 mg/dL was 16.14 mg/dL/min per day. The incidence of events >250 mg/dL was 0.14 per patient. Patients were exposed to an interstitial glucose value over 250 mg/dL in 6.59% of time and AUC >250 mg/dL was 1.66 mg/dL/min per day. We described progressive reduction of the percentage of time >180 mg/dl during the follow-up which became statistically significant on day 3 and 4. Because of the low number of patients who completed 5 days of follow-up, this tendency didn’t maintain on day 5. Similar tendency was described among the events >250 mg/dl without statistically significant difference.

Table 5.

Incidence of Events, Percentage on Time (%t), and AUC >180 mg/dL and >250 mg/dL.

	Events >180 mg/dL						Events >250 mg/dL
	Incidencea	P	%t	P	AUCb	P	Incidencea	P	%t	P	AUCb	P
1	2.23		43.68		21.20		1.15		10.37		3.20
2	2.35	.675	42.32	.705	16.60	.055	0.91	.262	7.51	.177	1.38	.016
3	2.35	.675	38.72	.166	15.36	.015	0.82	.123	6.42	.063	1.43	.019
4	1.96	.399	35.21	.043	14.66	.019	0.68	.056	6.20	.090	1.64	.074
5	1.86	.327	31.09	.011	11.57	.004	0.51	.028	4.04	.300	0.91	.026
6	2.28	.940	29.28	.087	7.47	.015	0.19	.051	1.47	.072	0.34	.102

^aEvents per patient.

^bmg/dL/min per day.

Discussion

Hyperglycemia in the hospital setting is associated with a higher rate of complications, longer hospital stay, costs and mortality.^1-5 Basal bolus insulin regimen allows optimal glycemic control in non–critically ill patients; however, it is associated with a risk of hypoglycemia. The use of CGM in hospitalized patients provides an accurate estimation of blood glucose levels and is more effective than POC in detection of hypoglycemic episodes.¹⁸ A recent consensus statement supports the use of CGM in the non-ICU setting in insulin treated patients with diabetes.²³ Our results highlight the efficacy of the basal bolus insulin regimen in the management of hospitalized patients with T2D. We found a time on targets of 87.2% at the end of the study, with a progressive reduction of interstitial glycemic average through hospitalization. In agreement with our results, Schaupp et al found a mean percentage of time on glycemic target of 75.8%, the first day 67.7% with a progressive increase to 77.5% (P < .04) the last day.¹⁴ The difference of the percentage of time on targets reached between the studies could be explained by the different definition of glycemic target and inclusion of patients with systemic steroid treatment, which could increase insulin resistance.

Our group previously reported that in patients treated with a basal bolus insulin regimen, CGM detected a higher number of hypoglycemic events than POC without differences in average daily glucose levels between CGM and POC.¹⁸ Burt et al conducted a study in patients with T2D who underwent CGM for 72 hours (System Gold Medtronic MiniMed, Northridge, CA) and POC testing. Ten out of 26 patients developed hypoglycemia <72.6 mg/dL (4 mmol/L) by CGM, but only one of these episodes was detected by POC.¹³ Similarly, Schaupp et al evaluated 84 patients with T2D, CGM detected a 15-fold increase in nocturnal hypoglycemia compared to capillary glucose testing.¹²

Other studies have assessed the safety of basal bolus insulin regimen using CGM.^13,14 We found similar rates of hypoglycemia with an optimal time on targets according to ADA 2018. However, in our study we found that mean glucose level and high glycemic variability were significant and independent factors that contributed to hypoglycemia (<70 mg/dl). Recently, Uemura et al performed a CGM-based retrospective study to identify the factors associated with hypoglycemia (<70 mg/dL) in hospitalized patients with T2D. Similar findings were described: lower fasting mean glucose values compared to the no-hypoglycemia group using ROC curve analysis showed a threshold value in the mean glucose of 150.9 mg/dL (AUC = 0.718, 95% CI 0.557, 0.880) and a SD of 41.1 mg/dL (AUC = 0.777, 95% CI 0.646, 0.907) to predict hypoglycemia.²⁴ Lichtenegger et al recently proved the efficacy and safety of insulin glargine U300 and an algorithm to calculate the dose of insulin (GlucoTab) in non–critically ill hospitalized patients of tertiary care by using CGM, achieving a time in target of 54.3% with decreased of hypoglycemia <70 mg/dl with 8 hypoglycemic episodes in 5 patients, only one severe. In this study, preexisting insulin therapy was a risk factor for hypoglycemia and a reduction in glycemic variability was observed.²⁵

GV has become a new target in T2D treatment in ambulatory setting to reduce hypoglycemia. Our findings suggest that reducing GV in hospitalized patient with T2D should be another target in the clinical setting. Clinical studies have shown that GV is a predictor of both hypo and hyperglycemia, independently of A1c levels, and the risk of asymptomatic hypoglycemia is directly related.^26,27 Different metrics have been associated as independent risk factors for diabetes complications in hospitalized patients, particularly cardiovascular disease, changes in cognitive function, postsurgery complications and increased mortality in ICU.^21,28 Even so, there are few clinical trials assessing GV in hospitalized patient, and more data should be analyzed to determine its relevance in hospital setting.

The strengths of this study are the use of the current definitions of the ADA for hypoglycemia and glycemic targets for hospitalized non–critically ill patients as well as the use of the CGM which allows evaluation of GV, a more accurate detection of hypo and hyperglycemia than POC testing.^11,18 Another strong point of this study is the report of the data obtained from CGM following the actual recommendations proposed by the ATTD international consensus on use of continuous glucose monitoring.²¹ Mean length of stay of our patients was longer in comparison with similar studies, 15 days ±11, due to the relocation in a high complexity hospital, but we can afford to evaluate only a period of 3 to 6 days of CGM per patient. The main limitation of this study is the lack of a control group to compare the basal bolus insulin regimen with the actual therapeutic alternatives to manage hyperglycemia in the hospital setting, such as DPP4 inhibitors and other insulin agents and the short follow-up time with CGM, which could explain the few events <54 mg/dL and <70 mg/dL detected. This few events of hypoglycemia <54 mg/dL is the reason that did not allow us to calculate the confidence intervals of hypoglycemia rates precisely.

Conclusion

Basal bolus insulin regimen is safe and effective for the management of non–critically ill T2D patients hospitalized in general wards. Our data show more than 80% of time in glycemic targets without an increase of the risk of clinically significant hypoglycemia (<54 mg/dL). Factors such as decrease of mean glucose level, and increased of glycemic variability parameters (CV%, SD) were associated to events below 70 mg/dL. This information could be useful to detect patients at high risk of hypoglycemia in the hospital setting. Future clinical trials that assess factors associated to hypoglycemia are required.