Introduction
The number of maintenance hemodialysis patients with diabetes has recently increased worldwide [
1]. Previous studies revealed that poor glycemic control causes high cardiovascular morbidity and mortality in these patients, but also that intensive diabetes therapy has long-term beneficial effects on the risk of cardiovascular disease and all-cause death [
2‐
5]. However, since the ACCORD study showed that severe hypoglycemia could increase the risk of cardiovascular death in participants with high underlying cardiovascular disease risk [
6], hypoglycemia has become one of the most important considerations in diabetes treatment. Sun et al. found that 54 of 102 type 2 diabetes patients on hemodialysis had symptomatic hypoglycemia during a 3-month follow-up period [
7]. However, asymptomatic hemodialysis-associated hypoglycemia might also be important in these patients. Patients with end-stage renal disease are at high risk of asymptomatic hypoglycemia [
8], and asymptomatic hemodialysis-associated hypoglycemia has been reported in patients with type 2 diabetes [
9‐
11].
Furthermore, high blood glucose (BG) variability in chronic glycemic control [standard deviation (SD) of glycated hemoglobin (HbA1c)] is associated with a high risk of hypoglycemia-related hospitalization [
12]. BG fluctuations activate oxidative stress and have been associated with arteriosclerosis [
13,
14]. The mean amplitude of glycemic exclusion (MAGE) is also used as an index of BG variability and has been reported to be associated with the presence and severity of coronary artery disease in patients with type 2 diabetes [
15‐
17]. Hypoglycemia can also lead to vascular injury and is associated with an adverse prognosis in patients with diabetes and chronic kidney disease [
18‐
21]. However, the precise BG profile and its variability in type 2 diabetes patients undergoing maintenance hemodialysis are unclear.
Continuous glucose monitoring (CGM) is useful for elucidating BG variability in diabetes patients by analyzing the amplitude and timing of glucose fluctuations [
22]. CGM is now commonly used for monitoring BG levels in patients undergoing maintenance hemodialysis [
9,
10,
23‐
32]. Nonetheless, there are only a few detailed reports on BG variability during hemodialysis therapy. Even though these studies showed a decrease in BG during hemodialysis [
9,
29,
31,
32], BG variability in these patients was complex, and its details are still unknown. Thus, further study involving CGM is needed to understand BG variability in diabetes patients undergoing maintenance hemodialysis.
Dipeptidyl peptidase-4 (DPP-4) inhibitors enhance the therapeutic effects of glucagon-like peptide-1 (GLP-1) by blocking its rapid degradation. Therefore, by increasing the circulating levels of biologically active GLP-1, DPP-4 inhibitors can elevate insulin levels. Furthermore, GLP-1 reduces meal-stimulated glucagon levels and improves BG control in diabetic patients [
33‐
36]. Some reports have suggested that DPP-4 inhibitors can potentially suppress BG variability in patients with type 2 diabetes [
37,
38]. Moreover, DPP-4 inhibitors can be safely used even in type 2 diabetes patients undergoing maintenance hemodialysis, in conjunction with appropriate dose regulation [
39,
40]. However, the effects of DPP-4 inhibitors on BG variability during hemodialysis and at nighttime in hemodialysis patients are unclear.
Accordingly, this study aimed to evaluate BG variability in type 2 diabetes patients undergoing maintenance hemodialysis by using CGM to assess the effectiveness of DPP-4 inhibitors, particularly during hemodialysis sessions and the subsequent nocturnal period.
Methods
Participants and Study Design
The primary outcome in this multicenter prospective observational exploratory study was BG variability in type 2 diabetes patients undergoing maintenance hemodialysis with or without a DPP-4 inhibitor [DPP-4 inhibitor (+) and DPP-4 inhibitor (−), respectively]. This variability was evaluated using the SD, MAGE, and a linear mixed-effects model. The frequency of hypoglycemia and the mean BG level were also evaluated. Participants were enrolled from November 2012 to March 2014 by T.I., M.H., and E.K. at Niigata University, Itoigawa General Hospital and Nagaoka Chuo Hospital, Niigata, Japan. The inclusion criteria were (1) age ≥ 20 years, (2) diagnosis of type 2 diabetes mellitus, (3) maintenance hemodialysis, and (4) the ability to provide informed consent. The exclusion criteria were (1) severe heart disease (New York Heart Association III or IV), (2) severe hepatic insufficiency, (3) apparent signs of current systemic infection or sepsis requiring active use of intravenous antibiotics, (4) perioperative status, (5) insulin therapy, (6) fasting, and (7) allergy to DPP-4 inhibitors. The reasons for hospitalization were shunt occlusions, pneumonia, and urinary tract infections. All participants were able to consume the prescribed meal, which had restricted calories, salt, and protein according to body weight, at 8:00 am, 12:00 pm, and 6:00 pm on the non-hemodialysis day and at 8:00 am, 2:00–3:00 pm after dialysis, and 6:00 pm on the day of hemodialysis.
This study was approved by the institutional review boards of Niigata University, Itoigawa General Hospital, and Nagaoka Chuo Hospital, and was performed in accordance with the principles embodied in the Declaration of Helsinki. All participants provided written informed consent. The Ethics Committee of the Niigata University School of Medicine approved the study (approval numbers: 2015-1476 and 2015-2119). The study was registered with the University Hospital Medical Information Network (UMIN) Clinical Trials Registry (UMIN000012638).
Measurement of Glucose Profiles
Glucose levels were measured by CGM for 48–168 h. A CGM System Gold (Medtronic MiniMed, Northridge, CA, USA) was used for 7 patients and a CGM System iPro2 (Medtronic MiniMed) was used for 38 patients. Patients underwent 3–5 h of hemodialysis 3 days per week; the dialysate glucose concentration was 100 mg/dL or 150 mg/dL. The CGM system was calibrated by nurses four times a day by comparison with the capillary BG value using the Medisafe-Mini (Terumo, Tokyo, Japan), a self-monitoring BG device. CGM was used to evaluate glucose levels from 7:00 am on the day of hemodialysis for 24 h. Hypoglycemia was defined as a CGM reading of < 70 mg/dL. The CGM test was performed for at least 2 weeks after DPP-4 inhibitor therapy was started. Pre-hemodialysis venous blood samples were obtained at the beginning of the week, and routine biochemical parameters, including HbA1c, were measured.
Statistical Analysis
For continuous variables, we used an unpaired t test for parametric variables and the Mann–Whitney U test for nonparametric variables. Differences in proportions were evaluated using Fisher’s exact test. Relationships between two continuous variables were evaluated using Spearman’s rank correlation coefficient. Mean SD and MAGE were calculated to evaluate BG variability. We calculated MAGE as the average of the variability above 1 SD of the 24 h wave of BG variability from 7:00 am to the next morning on the hemodialysis and non-hemodialysis days. A linear mixed-effects model was used to evaluate the between-group and within-group differences in the slope of the BG levels. The slope represents the estimated coefficient for changes in BG levels per unit time calculated using the linear mixed-effects model. The model included the time since initiation of hemodialysis and the time at night, the DPP-4 inhibitor (+) and (−) groups, and the interaction between time and treatment. The approximate inference regarding fixed or covariate effects in linear mixed-effects models was determined using the Kenward–Roger approximation method for degrees of freedom. Given that the study was exploratory in nature, the sample size was calculated to detect a difference in slope of 0.1 mg/dL per minute during a hemodialysis session with 70% power at the 5% level of significance in the analysis using the linear mixed-effects model. From this calculation, the number of cases required in each group was 14. All tests were two-sided and P values of less than 0.05 were considered statistically significant. The statistical analysis was performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and IBM SPSS Statistics 25 for Windows (SPSS Inc., Chicago, IL, USA).
Discussion
There is little information on BG variability in maintenance hemodialysis patients. Our study shows that maintenance hemodialysis patients with type 2 diabetes have high BG variability, as evidenced by the higher SD of BG, MAGE, and linear mixed-effects model coefficients. Also, our data from the linear mixed-effects model showed a rapid BG drop during hemodialysis and a tendency for hypoglycemia in the subsequent nighttime period in maintenance hemodialysis patients with type 2 diabetes. Although DPP-4 inhibitors have been shown to improve BG variability in patients with type 2 diabetes in many randomized controlled trials (RCTs) [
41‐
49] (see the table in the Electronic supplementary material), few studies have reported the same effect in maintenance hemodialysis patients [
50]. Our study illustrates the usefulness of DPP-4 inhibitors in these patients. Moreover, we showed that DPP-4 inhibitors ameliorated not only the 24 h BG variability but also the BG drop during hemodialysis and the tendency for nocturnal hypoglycemia in particular.
The ability of DPP-4 inhibitors to suppress BG variability in patients with type 2 diabetes has been shown in studies using CGM [
51,
52]. Furthermore, several RCTs using CGM found that BG variability is suppressed more effectively by DPP-4 inhibitors than by other agents such as sulfonylureas [
41,
42] and sodium glucose cotransporter 2 inhibitors [
43‐
45] or by combination with insulin therapy [
46‐
49]. The patients in these RCTs who showed suppressed BG variability with DPP-4 inhibitors had a mean age of less than 60 years [
42‐
44,
46,
48,
49], had HbA1c > 7% [
41‐
44,
46‐
49], were drug naïve [
41,
49], or had used metformin only [
41‐
43,
46]. There have also been some reports on the usefulness of DPP-4 inhibitors in patients on hemodialysis [
50], but none has shown obvious suppression of BG variability. In this prospective observational exploratory study, analysis using a linear mixed-effects model showed that DPP-4 inhibitors could suppress BG variability in patients with type 2 diabetes. However, it is not known whether DPP-4 inhibitors are more effective at suppressing BG variability than any other drugs, even in patients on hemodialysis. This is an issue to be resolved in the future. Several reports have examined the efficacy of DPP-4 inhibitors for suppressing BG variability not only during the day and after a meal but also at night [
53,
54]. However, until now, there has been no clear evidence that DPP-4 has a nocturnal effect. Using a linear mixed-effects model, we have shown that DPP-4 inhibitors could potentially suppress BG variability during hemodialysis sessions as well as subsequent nocturnal changes. Whether this effect is a feature of DPP-4 inhibitors in general or specific to patients on dialysis needs further investigation.
One of the essential endpoints of our study was the slope of the BG levels measured in mg/dL/min throughout hemodialysis and at night. This slope was calculated using CGM data obtained at baseline and every 5 min during the treatment phase. The difference in slope between the DPP-4 inhibitor (+) and (−) groups was evaluated with a linear mixed-effects model. CGM data usually show a periodic waveform. However, the changes between time points can be approximated using the slope of a linear mixed-effects model. Linear mixed-effects methods are used to fit the model to repeated-measures data such as those of CGM [
55], in which measurements are obtained repeatedly over time or under different conditions. The model comprises repeated effects, fixed or covariate effects, and random effects, and the interactions among combinations of these as predictor variables. In general, many study designs generate longitudinal or repeated-measures data sets. These designs are applied in a variety of settings throughout the medical, biological, and physical sciences. Repeated-measures data obtained using such a design often involve missing values, but the linear mixed-effects model can nonetheless estimate the parameters without ad hoc imputation of the missing values. Therefore, linear mixed-effects models provide researchers with powerful and flexible analytic tools for these types of data. They are considered especially useful for analysis of CGM data over time, such as during hemodialysis or at nighttime.
A comparison of our maintenance hemodialysis patients treated with and without DPP-4 inhibitors revealed that the average BG was similar but that the SD of BG and MAGE were significantly lower in the DPP-4 inhibitor (+) group. Some studies have shown that DPP-4 inhibitors, which can increase insulin secretion but reduce prandial glucagon, improve the BG wave [
51,
56]. The reduction in prandial glucagon is believed to be the most important mechanism through which DPP-4 inhibitors improve the BG wave [
35]. The present study showed that DPP-4 inhibitors are as effective in dialysis patients as in type 2 diabetes patients not undergoing hemodialysis [
41,
44,
48]. The mechanisms are unclear, and more detailed studies are needed, including those that analyze the fluctuation in glucagon.
The average BG, MAGE, and SD of BG were not significantly different between nonhemodialysis and hemodialysis days. However, a significant BG drop was seen during hemodialysis on the hemodialysis day. Asymptomatic hypoglycemia has been reported in hemodialysis patients during hemodialysis [
25,
57,
58]. It has long been thought that the diffusion of glucose from blood to dialysate causes this hypoglycemia, and that the use of glucose-containing dialysate can prevent such hypoglycemia [
11,
57,
59]. On the other hand, some patients experience a lower BG than the dialysate glucose concentration [
58], and hypoglycemia is not completely preventable by glucose-containing dialysate [
59]. In addition to chronic undernutrition [
60,
61], a delay in lunchtime due to hemodialysis extends the fasting time, causing a loss of glycogen in maintenance hemodialysis patients and increasing the likelihood of developing hypoglycemia. Other mechanisms for hypoglycemia in maintenance hemodialysis patients have been reported. Takahashi et al. stated that excessive consumption of glucose is the result of an accelerated anaerobic metabolism, and causes hypoglycemia during hemodialysis in these patients [
62].
Additionally, diabetic patients have less protection against a rapid fall in BG. Counter-regulation of the autonomic nervous system and the hormonal response to hypoglycemia are impaired by diabetic complications in patients with long-term diabetes [
11,
63,
64]. These conditions may impair the body’s ability to control BG and cause a loss of response to increasing BG in the hypoglycemic state. In the present CGM study, there were only a few cases of hypoglycemia during or after hemodialysis, but almost all patients experienced a rapid fall in BG during hemodialysis. Compared with the DPP-4 inhibitor (−) group, the rate of decrease in BG during hemodialysis was reduced in the DPP-4 inhibitor (+) group. Our study thus suggests that DPP-4 inhibitors can prevent a rapid BG drop during hemodialysis. The mechanisms are unclear, but DPP-4 inhibitors may be able to improve the fluctuation in BG during hemodialysis in a similar manner to their ability to ameliorate the circadian variation of BG [
51,
56,
65]. In terms of other factors, glucagon is probably associated with an improvement in BG fluctuation during hemodialysis in the same way as during nocturnal BG control. A previous study revealed that vildagliptin, a DPP-4 inhibitor, reduces postprandial glucagon levels and improves hyperglycemia in patients with type 2 diabetes [
35]. Also, DPP-4 inhibitors improve the ability of both β-cells and α-cells to sense and appropriately respond to hypoglycemia [
66]. Thus, DPP-4 inhibitors may be able to improve the response to hypoglycemia and prevent a rapid BG drop during hemodialysis in diabetic hemodialysis patients.
In our study, patients in the DPP-4 inhibitor (+) group seldom had BG < 70 mg/dL during the night. Diabetic patients frequently experience hypoglycemia due to a reduced autonomic nerve response, hormonal antagonism [
63], and glucagon secretion in response to hypoglycemia [
64]. Previous work also showed the potential ability of DPP-4 inhibitors to improve not only hyperglycemia but also hypoglycemia [
37,
38]. Excessive glucagon secretion was seen in diabetes patients as well as a lack of glucagon responsiveness to hypoglycemia. When constant oversecretion of glucagon is inhibited by DPP-4 inhibitors [
35], glucagon shows its normal response to a BG fall and may be able to prevent hypoglycemia [
66]. Hypoglycemia in hemodialysis patients can occur even when they have never been diagnosed with diabetes or have not been exposed to hypoglycemic agents [
25,
67]. Chronic kidney disease or hemodialysis can probably cause hypoglycemia per se [
7,
8,
11,
18,
62]. Glycogenesis is as vital in the kidney as it is in the liver [
68], especially under hypoglycemic conditions or during fasting [
68‐
74]. When BG falls, glucagon and catecholamine secretion boosts glycogenesis in the kidney and increases BG levels in healthy people [
75]. This mechanism is lost in maintenance hemodialysis patients, which is why they easily develop hypoglycemia under nocturnal fasting conditions.
Moreover, maintenance hemodialysis patients can often be undernourished due to protein loss during hemodialysis, and they often consume much smaller meals than recommended. Thus, not only are chronic malnutrition and lack of kidney glycogenesis important reasons for hypoglycemia in maintenance hemodialysis patients [
61], but energy loss during hemodialysis and reduced energy storage can also explain nocturnal hypoglycemia on hemodialysis days [
76]. Previous work showed that DPP-4 inhibitors improve hyperglycemia without inducing hypoglycemia during Ramadan, which has a long fasting time [
77]. This efficacy of DPP-4 inhibitors during Ramadan may be somewhat similar to their efficacy during nocturnal fasting in maintenance hemodialysis patients.
There are some limitations to this study. Firstly, this study had a small number of patients. Moreover, the subjects were hospitalized patients, who are evidently in a different setting from outpatients. Secondly, this was not a RCT, so there could be confounding. Thirdly, in this study, five types of DPP-4 inhibitors were prescribed for 15 patients. However, Craddy et al. [
78] recently reported that DPP-4 inhibitors have equivalent effects across the entire class in terms of essential efficacy and safety outcomes. Fourthly, although we surmise that glucagon plays a key role in hypoglycemia episodes and the prevention of hypoglycemia by DPP-4 inhibitors, we did not measure glucagon. This should be the next step. In the future, the use of only one type of DPP-4 inhibitor in a larger RCT is warranted.
Acknowledgements
Disclosures
Tomomi Ishikawa-Tanaka has received lecture fees from Daiichi Sankyo Co., Eli Lilly Japan K.K., Kyowa Hakko Kirin Co., Ltd., MSD K.K., Novartis Pharma K.K., Ono Pharmaceutical Co., Ltd., and Takeda Pharmaceutical Co., Ltd. Michihiro Hosojima and Akihiko Saito have received lecture fees from Daiichi Sankyo Co., Ltd., Eli Lilly Japan K.K., Kowa Company, Ltd., Kyowa Hakko Kirin Co., Ltd., MSD K.K., Mitsubishi Tanabe Pharma Corporation, Nippon Boehringer Ingelheim Co., Ltd., Novartis Pharma K.K., Ono Pharmaceutical Co., Ltd., and Takeda Pharmaceutical Co., Ltd. Hideyuki Kabasawa has received lecture fees from Eli Lilly Japan K.K., Kyowa Hakko Kirin Co., Ltd., MSD K.K., Mitsubishi Tanabe Pharma Corporation, Nippon Boehringer Ingelheim Co., Ltd., and Ono Pharmaceutical Co., Ltd. Ryohei Kaseda has received lecture fees from Daiichi Sankyo Co., Ltd., Kyowa Hakko Kirin Co., Ltd., MSD K.K., Nippon Boehringer Ingelheim Co., Ltd., Novartis Pharma K.K., Sanwa Kagaku Kenkyusyo Co., Ltd., and Takeda Pharmaceutical Co., Ltd. Ryota Yasukawa has received lecture fees from Kyowa Hakko Kirin Co. and Mitsubishi Tanabe Pharma Corporation. S.K. has received a lecture fee from MSD K.K. Emiko Kono has received a lecture fee from Kyowa Hakko Kirin Co. Noriaki Iino has received lecture fees from Daiichi Sankyo Co., Ltd., Kyowa Hakko Kirin Co., Ltd., MSD K.K., Mitsubishi Tanabe Pharma Corporation, Ono Pharmaceutical Co., Ltd., Sanwa Kagaku Kenkyusyo Co., Ltd., and Takeda Pharmaceutical Co., Ltd. Yoshiki Suzuki has received lecture fees from Daiichi Sankyo Co., Ltd., Kowa Company, Ltd, Kyowa Hakko Kirin Co., Ltd., and Mitsubishi Tanabe Pharma Corporation. Ichiei Narita has received lecture fees from Kowa Company, Ltd. and Kyowa Hakko Kirin Co. Michihiro Hosojima and Akihiko Saito have received donations for research from Daiichi Sankyo Co., Ltd., Kowa Company, Ltd., Kyowa Hakko Kirin Co., Ltd., MSD K.K., Mitsubishi Tanabe Pharma Corporation, Sanwa Kagaku Kenkyusyo Co., Ltd., and Takeda Pharmaceutical Co., Ltd. Akihiko Saito has also received donations for research from Eli Lilly Japan K.K., Nippon Boehringer Ingelheim Co., Ltd., Novartis Pharma K.K., and Ono Pharmaceutical Co., Ltd. Akihiko Saito has received a collaborative research grant from Kyowa Hakko Kirin Co. Shoji Kuwahara has received donations for research from MSD K.K. Noriaki Iino has received donations for research from Kyowa Hakko Kirin Co., Ltd. and Takeda Pharmaceutical Co., Ltd. Ichiei Narita has received research support from Daiichi Sankyo Co., Ltd., Kyowa Hakko Kirin Co., Ltd., and Ono Pharmaceutical Co. Yusuke Yata, Takuma Takata, Takahiro Tanaka and Nobutaka Kitamura have nothing to disclose.