Fasting glucose level is associated with nocturnal hypoglycemia in elderly male patients with type 2 diabetes

Abstract

Background: Nocturnal hypoglycemia was a common and serious problem among patients with type 2 diabetes (T2DM), especially in the elderly. This study investigated whether fasting glucose was an indicator of nocturnal hypoglycemia in elderly male patients with T2DM.

Methods: A total of 291 elderly male type 2 diabetic patients who received continuous glucose monitoring (CGM) between January 2007 and January 2011 were enrolled in the study. The association of fasting glucose and nocturnal hypoglycemia based on CGM data was analyzed, comparing with bedtime glucose.

Results: Based on CGM data, patients with nocturnal hypoglycemia had significantly lower fasting glucose (5.88 ± 1.29 versus 6.92 ± 1.32 mmol/L) and bedtime glucose (7.33 ± 1.70 versus 8.01 ± 1.95 mmol/L) than patients without nocturnal hypoglycemia (both p < 0.01). Compared with the highest quartile, the lowest quartile of fasting glucose had a significantly increased risk of nocturnal hypoglycemia after the multiple adjustments (p_{for trend} < 0.001). However, this association did not appear in bedtime glucose. When the prediction of nocturnal hypoglycemia either by fasting glucose or bedtime glucose using the area under receiver operating characteristic (ROC) curve, fasting glucose but not bedtime glucose, was an indicator of nocturnal hypoglycemia, with an area under the ROC curve (AUC) of 0.714 (95% CI: 0.653 ∼ 0.774, p < 0.001). On the ROC curve, the Youden index was maximal when fasting glucose was 6.1 mmol/L.

Conclusions: Fasting glucose may be a convenient and clinically useful indicator of nocturnal hypoglycemia in elderly male patients with T2DM. Risk of nocturnal hypoglycemia significantly increased when fasting glucose was less than 6.1 mmol/L.

• Fasting glucose
• nocturnal hypoglycemia
• type 2 diabetes

Introduction

Glycemic management of diabetes was limited by the barrier of nocturnal hypoglycemia. Risk of hypoglycemia should be considered for every patient with type 2 diabetes (T2DM) before intensifying the therapeutic regimen, especially in the elderly. Aging was associated with an increasing risk of hypoglycemia, and hypoglycemia significantly contributed to cardiovascular morbidity and mortality in the elderly [1]. Continuous glucose monitoring (CGM) system is a useful tool to diagnose asymptomatic nocturnal hypoglycemia. However, CGM system cannot be widely adopted in the clinical practice due to its expensive sensors. Therefore, various variables predicting risk of nocturnal hypoglycemia have been tested. Pena et al. [2] indicated that hypoglycemia evaluated as low blood glucose index (LBGI) was related to vascular function in children with type 1 diabetes. Data from PREDICTIVE suggested fasting glucose variability could serve as a useful marker for risk of nocturnal hypoglycemia in clinical practice [3]. Nevertheless, LBGI and fasting glucose variability were indicators calculated or monitored complicatedly. Several studies tried to explore a new convenient indicator of nocturnal hypoglycemia in various ages. Bedtime blood glucose was clarified to be a predictor of nocturnal hypoglycemia in children with type 1 diabetes [4]. In the study by Garcia-Patterson et al. [5], pregestational BMI was demonstrated as a predictor of nocturnal hypoglycemia in offspring of women with gestational diabetes mellitus. In addition, some studies have tried to predict nocturnal hypoglycemia from electroencephalograph (EEG) changes [6,7]. However, it is still unknown if these indicators pertain to elderly patients with T2DM. This study investigated whether fasting glucose was an indicator of nocturnal hypoglycemia in elderly male patients with T2DM.

Methods

Study subjects

This study was conducted as a cross-sectional study, which enrolled 291 elderly male type 2 diabetic in-patients who received continuous glucose monitoring (CGM) between January 2007 and January 2011 in China. The inclusion criteria comprised a diagnosis of type 2 diabetes, male, aged ≥60 years, and treated with diet, oral hypoglycemic agents or insulin. Exclusive criteria were changes in lifestyle, diet or treatment and acute illness within 1 month before receiving CGM, or concomitant medications that may affect glucose metabolism, such as glucocorticoids and thyroid hormones. Data collection and analysis of the study subjects were approved by the local ethics committee, and the study was conducted in accordance with the Declaration of Helsinki.

Study design

A standardized questionnaire was adopted by trained data collectors to obtain information on demographic characteristics, personal and family medical history, physical examination, laboratory test outcomes, management and data of CGM. The personal medical history consisted of the diagnosis and treatment of diabetes, diabetes complication, hypertension, dyslipidemia and cardiovascular events. Physical examination included height, body weight, waist circumference (WC), systolic blood pressure (SBP) and diastolic blood pressure (DBP). Laboratory test comprised glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), postprandial plasma glucose (2h-PPG), total cholesterol (TC), triglycerides (TG), low density lipoprotein-cholesterol (LDL-c) and high density lipoprotein-cholesterol (HDL-c). Plasma glucose was measured with a glucose oxidase method. Serum TC, TG, LDL-c and HDL-c were determined with enzymatic methods using Cobas® 8000 system (Roche Diagnostics).

Continuous glucose monitoring (CGM)

All patients received 24-hour Medtronic MiniMed® Continuous Glucose Monitoring System™ (CGMS; Medtronic MiniMed, Northridge, CA). The CGM system sensor (Medtronic, Northridge, CA) was inserted into all subjects by the same specialized technician during 15:00–17:00 in hospital and calibrated according to the standard operating guidelines for 72 h before removed. The exercise and calorie intake of subjects under investigation were similar to those at home, but with relatively disciplinary dietary time in hospital, including 07:15 A.M. for breakfast, 11:15 A.M. for lunch and 5:15 P.M. for dinner. MiniMed Solutions Sensor™ 3.0 software (Medtronic, Northridge, CA, USA) package was used to download data. The manufacturer’s accuracy criteria were applied: a correlation between the sensor and meter readings of at least 0.79 and a mean absolute difference of at most 28% [8]. Data not meeting these criteria were excluded. Nocturnal hypoglycemia was defined as any sensor value <3.9 mmol/L [9,10] during 0:00–8:00 of CGM. The association of either fasting glucose or bedtime glucose with nocturnal hypoglycemia based on CGM data was analyzed and compared.

Data analyses

Data were presented as mean ± standard deviation (SD), unless otherwise stated. Subjects were divided into two groups based on patients experiencing nocturnal hypoglycemia or not. Comparison of variables among groups was performed using unpaired Student’s t test or χ² test. The independent significant risk factor of nocturnal hypoglycemia was evaluated by multivariate logistic regression analysis with the calculation of the standardized odds ratios after adjustment by several relevant risk factors. In addition, areas under the receiver operating characteristics (ROC) curves were determined for each variable to identify the indicators of nocturnal hypoglycemia. The area under the curve (AUC) of ROC curves was a summary of the overall diagnostic accuracy of the test. p Values <0.05 were considered statistically significant.

Results

Clinical characteristics of participants

The prevalence of nocturnal hypoglycemia was 42.6% (124/291) in elderly male patients with T2DM during CGM. The mean duration of nocturnal hypoglycemia was 71.9 ± 77.8 min per day, with the lowest sensor value of 3.0 ± 0.6 mmol/L (2.2–3.8 mmol/L). The therapeutic strategies of the study subjects included 11 diet (3.8%), 115 oral hypoglycemic agents (39.5%) and 165 insulin (56.7%). The clinical characteristics of study subjects are shown in Table 1. FPG was lower in the group of subjects with nocturnal hypoglycemia than in the group of subjects without nocturnal hypoglycemia (p = 0.046). While no significant difference was found regarding to HbA1c and 2h-PPG. Other demographic and clinical characteristics such as age, BMI, WC, diabetes duration, SBP, DBP, TC, TG, LDL-c and HDL-c were comparable between the two groups (all p > 0.05).

Table 1. Clinical characteristics and therapeutic strategy of study subjects (mean±SD).

Variables	With nocturnal hypoglycemia (n = 124)	Without nocturnal hypoglycemia (n = 167)	p
Age (year)	75.7 ± 9.1	76.7 ± 9.1	0.365
BMI (kg/m^2)	25.2 ± 2.6	25.1 ± 2.7	0.843
WC (cm)	95.9 ± 8.6	94.4 ± 9.0	0.178
Diabetes duration (year)	16.7 ± 9.6	16.2 ± 9.7	0.629
SBP (mmHg)	128.4 ± 12.3	127.6 ± 11.6	0.598
DBP (mmHg)	72.1 ± 8.7	71.9 ± 8.0	0.860
HbA1c (%)	7.02 ± 1.09	7.21 ± 1.31	0.206
FPG (mmol/L)	5.88 ± 1.92	6.30 ± 1.59	0.046
2h-PPG (mmol/L)	9.32 ± 2.83	9.45 ± 2.81	0.702
TC (mmol/L)	4.11 ± 0.86	4.20 ± 0.81	0.380
TG (mmol/L)	1.38 ± 0.70	1.46 ± 0.97	0.426
LDL-c (mmol/L)	2.38 ± 0.72	2.40 ± 0.66	0.739
HDL-c (mmol/L)	1.13 ± 0.27	1.14 ± 0.36	0.787

BMI, body mass index; WC, waist circumference; SBP, systolic blood pressure; DBP, diastolic blood pressure; HbA1c, glycated hemoglobin; FPG, fasting plasma glucose; 2h-PPG, 2h plasma glucose; TC, total cholesterol; TG, triglyceride; LDL-c, LDL cholesterol; HDL-c, HDL cholesterol.

Data are the means ± SD.

Based on CGM data, subjects with nocturnal hypoglycemia had significantly lower fasting glucose (5.88 ± 1.29 versus 6.92 ± 1.32 mmol/L, p < 0.001) and bedtime glucose (7.33 ± 1.70 versus 8.01 ± 1.95 mmol/L, p = 0.002) than subjects without nocturnal hypoglycemia (Table 2). Mean blood glucose (MBG), either during daytime or nighttime, were lower in the group of subjects with nocturnal hypoglycemia than in the group of subjects without nocturnal hypoglycemia (both p < 0.01), while standard deviation blood glucose (SDBG) either during daytime or nighttime, were higher in the group of subjects with nocturnal hypoglycemia than in the group of subjects without nocturnal hypoglycemia (both p < 0.01).

Table 2. CGM variables of study subjects (mean ± SD).

Variables	With nocturnal hypoglycemia (n = 124)	Without nocturnal hypoglycemia (n = 167)	p
MBG (mmol/L)	7.11 ± 1.07	7.88 ± 1.45	<0.001
Daytime MBG (mmol/L)	7.79 ± 1.35	8.27 ± 1.62	0.007
Nighttime MBG (mmol/L)	5.94 ± 1.12	7.21 ± 1.42	<0.001
SDBG (mmol/L)	1.90 ± 0.77	1.43 ± 0.66	<0.001
Daytime SDBG (mmol/L)	1.68 ± 0.71	1.36 ± 0.65	<0.001
Nighttime SDBG (mmol/L)	0.92 ± 0.51	0.76 ± 0.48	0.006
Fasting glucose (mmol/L)	5.88 ± 1.29	6.92 ± 1.32	<0.001
Bedtime glucose (mmol/L)	7.33 ± 1.70	8.01 ± 1.95	0.002

MBG, mean blood glucose; SDBG, standard deviation of blood glucose.

Prevalence of nocturnal hypoglycemia categorized by quartiles of fasting glucose or bedtime glucose

Prevalence of nocturnal hypoglycemia categorized by quartiles of fasting glucose or bedtime glucose was analyzed. 70.4% patients with lowest quartile (quartile 1, fasting glucose <5.55 mmol/L) of fasting glucose had nocturnal hypoglycemia, which was significantly higher than other quartiles (quartile 2, 5.55 ∼ 6.30 mmol/L, 45.9%; quartile 3, 6.30 ∼ 7.25 mmol/L, 28.2%; quartile 4, ≥7.25 mmol/L, 26.7%; p < 0.001, Figure 1A). Successive quartiles of bedtime glucose had 58.0%, 39.2%, 42.5% and 31.9% of nocturnal hypoglycemia, respectively (p = 0.016, Figure 1B).

Figure 1. Prevalence of nocturnal hypoglycemia categorized by quartiles of fasting glucose (A) or bedtime glucose (B).

Association between fasting glucose or bedtime glucose and nocturnal hypoglycemia

The association between fasting glucose or bedtime glucose and nocturnal hypoglycemia was analyzed with multivariate logistic regression analysis (Table 3). Model 1 adjusted for age, BMI, WC and diabetic duration. Model 2 further adjusted for insulin, sulfonylureas or glinides therapy based on model 1, and model 3 further adjusted for HbA1c, FPG and 2h-PPG based on model 2. The lowest quartile of fasting glucose had a significantly increased risk of nocturnal hypoglycemia after the multiple adjustments (model 1, OR = 6.11; model 2, OR = 6.40; model 3, OR = 6.41; all p_{for trend} < 0.001) compared with the highest quartile. However, this association did not appear in bedtime glucose after multiple adjustments.

Table 3. Association between fasting glucose or bedtime glucose and nocturnal hypoglycemia.

	Unadjusted	Model 1	Model 2	Model 3
Fasting glucose
Quartile 1 (<5.6 mmol/L)	6.55 (3.18–13.5)	6.11 (2.71–13.8)	6.40 (2.73–15.0)	6.41 (2.57–16.0)
Quartile 2 (5.6–6.3 mmol/L)	2.34 (1.12–4.64)	1.65 (0.77–3.57)	1.78 (0.80–3.93)	1.72 (0.72–4.08)
Quartile 3 (6.3–7.3 mmol/L)	1.08 (0.52–2.23)	0.99 (0.45–2.18)	1.11 (0.50–2.50)	1.06 (0.46–2.46)
Quartile 4 (≥7.3 mIU/L)	1.00	1.00	1.00	1.00
p	<0.001	<0.001	<0.001	<0.001
Bedtime glucose
Quartile 1 (<6.4 mmol/L)	2.94 (1.48–5.85)	2.34 (1.05–5.20)	2.93 (1.23–6.94)	2.86 (1.18–6.95)
Quartile 2 (6.4–7.4 mmol/L)	1.37 (0.70–2.71)	1.23 (0.56–2.66)	1.53 (0.67–3.49)	1.61 (0.69–3.74)
Quartile 3 (7.4–8.6 mmol/L)	1.57 (0.80–3.10)	1.43 (0.68–3.02)	1.65 (0.76–3.57)	1.60 (0.72–3.56)
Quartile 4 (≥8.6 mmol/L)	1.00	1.00	1.00	1.00
p	0.018	0.176	0.099	0.138

Data are OR (95%CI). Participants with nocturnal hypoglycemia was defined as 1 and without as 0.

Model 1: Adjusted for age, BMI, waist circumference, and diabetic duration.

Model 2: Further adjusted for insulin, sulfonylureas or glinides therapy based on model 1.

Model 3: Further adjusted for HbA1c, fasting plasma glucose and postprandial plasma glucose based on model 2.

Area under the ROC Curve for the prediction nocturnal hypoglycemia by either fasting glucose or bedtime glucose

When the prediction of nocturnal hypoglycemia by either fasting glucose or bedtime glucose was compared using the area under receiver operating characteristic (ROC) curve, fasting glucose but not bedtime glucose was a predictor of nocturnal hypoglycemia, with an area under the ROC curve (AUC) of 0.714 (95%CI: 0.653 ∼ 0.774, p < 0.01) and 0.604 (95%CI: 0.538 ∼ 0.670, p = 0.002), respectively (Figure 2).

Figure 2. Area under the ROC curve for fasting glucose or bedtime glucose predicting nocturnal hypoglycemia. Using the area under receiver operating characteristic (ROC) curve, fasting glucose but not bedtime glucose, was a predictor of nocturnal hypoglycemia, with an area under the ROC curve (AUC) of 0.714 (95%CI: 0.653 ∼ 0.774, p < 0.001) and 0.604 (95%CI: 0.538∼0.670, p = 0.002).

To further analyze the optimal cutpoint of fasting glucose or bedtime glucose predicting abnormal nocturnal hypoglycemia (Table 4), the Youden index was maximal on a receiver operating characteristic curve where fasting glucose was 6.1 mmol/L (sensitivity: 61.3%; specificity: 70.7%) or bedtime glucose was 6.9 mmol/L (sensitivity: 46.8%; specificity: 68.9%).

Table 4. Optimal cutpoint of fasting glucose or bedtime glucose predicting nocturnal hypoglycemia.

	Fasting glucose			Bedtime glucose
Blood glucose (mmol/L)	Sensitivity	Specificity	Youden index	Sensitivity	Specificity	Youden index
5.7	42.7 (53)	83.8 (140)	0.265	15.3 (19)	91.6 (153)	0.069
5.9	52.4 (65)	76.0 (127)	0.284	20.2 (25)	88.6 (148)	0.088
6.1	61.3 (76)	70.7 (118)	0.320	22.6 (28)	86.2 (144)	0.088
6.3	67.7 (84)	63.5 (106)	0.312	30.6 (38)	83.2 (139)	0.138
6.5	72.6 (90)	58.7 (98)	0.313	35.5 (44)	77.2 (129)	0.127
6.7	76.6 (95)	52.1 (87)	0.287	37.9 (47)	72.5 (121)	0.104
6.9	79.0 (98)	47.9 (80)	0.269	46.8 (58)	68.9 (115)	0.157
7.1	83.9 (104)	38.9 (65)	0.228	50.0 (62)	62.3 (104)	0.123
7.3	84.7 (105)	31.1 (52)	0.158	53.2 (66)	56.3 (94)	0.095
7.5	87.1 (108)	25.7 (43)	0.128	57.3 (71)	52.1 (87)	0.094

Data are % (n).

Discussion

This study suggested that fasting glucose was a convenient and clinically useful indicator of nocturnal hypoglycemia in elderly male patients with T2DM. We found an independent association of fasting glucose and nocturnal hypoglycemia in elderly patients based on CGM data. These findings extend previous observations on exploring effective indicators of nocturnal hypoglycemia.

Our finding could have potential clinical relevance. Nocturnal hypoglycemia was a common and serious problem among patients with type 2 diabetes, especially in the elderly, because islet β cell function decreased gradually with aging and extension of diabetic duration [11]. Nocturnal hypoglycemia was a significant barrier of diabetes management. Prevention and elimination of nocturnal hypoglycemia from patients with diabetes was a critical component of diabetes management [12]. An effective variable predicting risk of nocturnal hypoglycemia may help to prevent nocturnal hypoglycemia. Our study demonstrated fasting glucose could be a convenient and clinically useful indicator to predict the risk of development of nocturnal hypoglycemia. If fasting glucose was less than 6.1 mmol/L in the elderly, increased risk of nocturnal hypoglycemia might be implied. Because the studied subjects received different treatment regimens, the risk of nocturnal hypoglycemia should be distinct. The insulin- and sulfonylurea-treated patients would be expected to have a higher risk of hypoglycemia. If subjects were only treated with metformin, they should not have as high a risk of developing hypoglycemia as the insulin-/sulfonylurea-treated subjects. In addition, subjects who received diet control only should not have risk of nocturnal hypoglycemia.

The mechanism of association between fasting glucose and nocturnal hypoglycemia remained unclear. Nocturnal hypoglycemia was associated with absolute or relative insulin excess, whether from injected or from secreted insulin [13]. Absolute insulin excess occurred when insulin or insulin secretagogue doses were excessive [14]. Relative insulin excess was related to increased insulin sensitivity in the middle night, or decreased insulin clearance in the state of renal failure [15]. It can be deduced that absolute or relative insulin excess was not only associated with nocturnal hypoglycemia but resulted in lower fasting glucose level. Other potential mechanisms of nocturnal hypoglycemia, such as defective glucose counterregulation, could be accounted for decreased fasting glucose level [16]. When deficiency of glucagon and epinephrine responses to hypoglycemia were demonstrated [17,18], especially in elderly patients with type 2 diabetes, risk of nocturnal hypoglycemia was markedly increased. The impairment of glucagon and epinephrine secretion, which decreased hepatic glucose production, might also have a corresponding influence on fasting glucose in the morning.

Hypoglycemic episodes were identified to be frequent in older adults with poor glycemic control; nevertheless, some data also suggested that raising HbA1c goals might not be adequate to prevent hypoglycemia in this population [19]. The present data confirmed that HbA1c level was similar between patients with and without nocturnal hypoglycemia. Previous studies have investigated various approaches to reduce risk of nocturnal hypoglycemia [20]. LBGI was proposed to provide an accurate assessment of nocturnal hypoglycemia risk [21]. However, LBGI was seldom selected as a predictor of nocturnal hypoglycemia risk in clinical practice mainly due to its complicated mathematical calculation. CGM system had been demonstrated to monitor nocturnal hypoglycemia effectively and associated with reduced time spent in hypoglycemia [22]. This study used CGM system to evaluate nocturnal hypoglycemia and suggested that fasting glucose was a convenient and clinically useful indicator of nocturnal hypoglycemia in patients with type 2 diabetes.

The principal limitation of this study was the cross-sectional study design and limited subject number. Therefore, our conclusions should be considered to be preliminary. Furthermore, the limited accuracy of the CGM device might influence the result. Episodes of plasma glucose less than 3.9 mmol/L were undoubtedly missed, and some low episodes identified were undoubtedly in error. In addition, the impact of Somogyi phenomenon on the fasting plasma glucose levels was not be considered in this study for limited data.

This study demonstrates that fasting glucose may be a convenient and clinically useful indicator of nocturnal hypoglycemia in elderly male patients with T2DM. Nocturnal hypoglycemia as measured with the CGM system might contribute to understand its relationship with FPG. Future prospective follow-up studies are needed to verify the association of FPG and nocturnal hypoglycemia.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.

Notes

¹The first two authors, Fusheng Fang and Haiying Xiao, contributed equally to this article.