1 Background

The anti-diabetic actions of dipeptidyl peptidase IV (DPP-4) inhibitors are based on the effects of two incretin hormones, glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP), which are rapidly inactivated by DPP-4. These incretins potentiate glucose-induced insulin release. In addition, GLP-1 also suppresses glucagon release and slows down gastric emptying during the postprandial period [1, 2]. Meta-analyses of randomized clinical trials have reported that DPP-4 inhibitors significantly improve the HbA1c levels in comparison with placebos, with an incidence of hypoglycemia similar to that observed with placebos [2–4]. These results are consistent with the glucose-dependent actions of incretins. However, severe hypoglycemia occurred in five patients treated with sitagliptin, either with monotherapy or in combination with metformin, in three different clinical trials [5–7]. Notably, although two of these trials did not report the events in the literature, information regarding severe hypoglycemia was described in the data for the drug application in the US, which can be retrieved from the FDA website. [2, 3, 5, 6]. In daily clinical practice, hypoglycemia-related emergency department and outpatient visits due to DPP-4 inhibitor monotherapy have been reported [8], and the number of patient-reported hypoglycemic episodes is two times higher with the use of sitagliptin than placebos [4].

Patient self-reports are more sensitive to underlying changes in functional status than health care providers’ reports and appear to be more suitable for assessing symptomatic adverse drug events (ADEs) [9, 10]. Non-serious symptomatic ADEs, which affect the real benefit-risk ratios of drugs, are often missed by health care providers. Recently, Hakobyan et al. [10] reported that health care provider-oriented assessment methods using medical record reviews offered considerably lower rates for gastrointestinal adverse drug reactions (ADRs) of oral glucose-lowering drugs than a patient questionnaire method. In 2003, the Japan Pharmaceutical Association (JPA) initiated a Drug Event Monitoring (DEM) project involving nationwide surveillance for ADEs using a patient questionnaire method in cooperation with JPA member pharmacies conducted in the fourth week of February every year [11]. The DEM project includes selected drugs or drug classes that have an impact on daily practice because of their targeted or adverse effects. DPP-4 inhibitors were the focus of investigation in 2012. The affiliated pharmacists put up a poster advertising the projects and noted the patients who were receiving the focused drug prior to the surveillance. Then during the 7-day surveillance period in the 4th week of February, every visitor who had been identified as a potential subject was asked to participate in the project. The aim of this study was to assess the profile of patient-reported symptomatic ADRs under DPP-4 inhibitor therapy and to detect risk factors for hypoglycemic and non-hypoglycemic adverse symptoms in daily clinical practice.

2 Methods

2.1 Study Design

Every patient who was receiving a DPP-4 inhibitor (sitagliptin, alogliptin, vildagliptin or linagliptin) from pharmacies affiliated with a cross-sectional nationwide survey of the JPA DEM project and visited there with a prescription of any drug(s) during the surveillance from 20 through 26 February 2012 were asked to participate in the project. The participants were interviewed by pharmacists using structured questionnaires to assess the self-perception of adverse symptoms after the last prescription of a DPP-4 inhibitor. The median duration from the last prescription to the interview was 28 (4–88) days. The questionnaires included open questions regarding demographic information and adverse symptoms, and a check list of symptoms that can be caused by DPP-4 inhibitors (Supplemental Table 1). The pharmacists asked participants further regarding their symptoms as needed and filled in the check list. A total of 649 (89.1 %) pharmacies in Kumamoto prefecture, Japan, responded to the survey, and the survey results of 2,002 participants were collected via a Web-based system linked to a reporting portal in the KUMAYAKU Network for Community Pharmacies in Kumamoto. This protocol was approved by the institutional ethics committees of the Faculty of Life Sciences, Kumamoto University.

2.2 Data Analysis

Patients with dementia or missing data of adverse symptoms, diseases under pharmacotherapy, alcohol consumption and/or smoking habits were excluded from the present study. Consequently, 1,550 patients were included in the statistical analyses. The patient characteristics, i.e. the demographic characteristics listed in Table 1, were not significantly different between the included and excluded patients (Supplemental Table 2). We defined hypoglycemic symptoms, on a theoretical basis, according to the review by Barnett et al. [12].

Table 1 Characteristics of the study population
Full size table

We identified independent risk factors associated with hypoglycemic or nonhypoglycemic symptoms during DPP-4 inhibitor use, employing univariate and multiple logistic regression analyses with calculation of the odds ratios (ORs) and 95 % confidence intervals (95 % CIs). In the multiple regression model, the ORs were adjusted for potentially confounding factors, including patient characteristics (e.g. age, sex) and factors that showed a statistical tendency (P < 0.1) in the univariate model. To assess the goodness of fit of our logistic regression models, we performed the Hosmer–Lemeshow test and calculated Nagelkerke’s R squared. In addition, we investigated the association between the daily dose and diseases or sex and assessed the effects on adverse symptoms. To assess the daily doses of four different DPP-4 inhibitors in the same way, we divided the prescribed daily dose for each subject by the defined maximum dose. The maximum daily doses approved for the treatment of diabetes mellitus (DM) in Japan were used as the defined maximum doses, i.e. 100 mg for sitagliptin and vildagliptin, 25 mg for alogliptin and 5 mg for linagliptin. The presence of renal or liver disease was defined as the use of pharmacotherapy for these diseases.

Categorical and continuous variables were also compared using Fisher’s exact test and Student’s t test, respectively. A P value of <0.05 was considered to be statistically significant. All statistical analyses were performed using the SPSS software package for Windows (Version 17.0, IBM Japan Ltd., Tokyo, Japan).

3 Results

The mean age of the subjects was 68.0 ± 11.8 years (19–98 years), including 864 males and 686 females (Table 1). The prescribed DPP-4 inhibitors included sitagliptin in 1,169 subjects, alogliptin in 241 subjects, vildagliptin in 136 subjects and linagliptin in 4 subjects (Table 1). Hypoglycemic symptoms occurring after previous prescriptions during a median interval of 28 days were reported by 34 (2.2 %) of 1,550 subjects: 13/532 (2.4 %) under monotherapy consisting of sitagliptin (10/402) or alogliptin (3/65) and 21/1,018 (2.1 %) under combination therapy consisting of sitagliptin (15/767) or alogliptin (6/176) and other hypoglycemic agents; however, no patients were treated with either vildagliptin or linagliptin (Supplemental Table 3). The most common symptoms were dizziness and hunger (0.9 and 0.4 %, respectively) as shown in the symptom profiles (Table 2).

Table 2 Reported symptoms
Full size table

In the univariate regression model, liver disease was found to be a risk factor for hypoglycemic symptoms (P = 0.032) (Supplemental Table 4). The type of DPP-4 inhibitor was not found to be associated with hypoglycemic or nonhypoglycemic symptoms (data not shown).

In the multiple regression model, hypoglycemic symptoms were found to be associated with liver disease (OR 3.17, 95 % CI 1.07–9.43), female sex (OR 3.26, 95 % CI 1.39–7.62) and alcohol consumption more than three times per week (OR 3.68, 95 % CI 1.30–10.43) (Table 3). When stratified according to the use of DPP-4 inhibitors as monotherapy or combination therapy, liver disease was identified as the sole independent risk factor (OR 5.11, 95 % CI 1.04–25.24) in patients receiving monotherapy, whereas female sex (OR 7.94, 95 % CI 2.42–26.10) and alcohol consumption of 1–3 times per week (OR 5.53, 95 % CI 1.31–23.40) or more than three times per week (OR 9.71, 95 % CI 2.55–36.90) were identified as risk factors in patients receiving combination therapy (Table 4).

Table 3 Crude and adjusted odds ratios for the symptoms of hypoglycemia
Full size table
Table 4 Crude and adjusted odds ratios for the symptoms of hypoglycemia among individuals receiving monotherapy or combination therapy with other DM medications
Full size table

Nonhypoglycemic symptoms were reported by 57 (3.7 %) individuals receiving monotherapy with sitagliptin (11/402) or vildagliptin (1/61) and combination therapy with sitagliptin (28/767), alogliptin (13/176) and vildagliptin (4/75) (Supplemental Table 3). Gastrointestinal symptoms (2.1 %) were the most common nonhypoglycemic symptom (Table 2). In the logistic regression analysis, combination therapy was identified as the sole risk factor for non-hypoglycemic symptoms (OR 1.97, 95 % CI 1.03–3.78) (Supplemental Table 5). The adjusted regression models were statistically fitted for hypoglycemic symptoms among all subjects, among individuals receiving monotherapy and among those receiving combination therapy, as well as for non-hypoglycemic symptoms among all subjects (P = 0.262, 0.412, 0.674 and 0.419, respectively, Hosmer–Lemeshow test). The other hand, the percentages explained by the variances (i.e. Nagelkerke’s R squared) were 4.5, 4.2, 9.4 and 2.0 %, respectively, for these regression models.

The daily dose did not differ between females and males; however the females took significantly higher doses per weight than the males (sitagliptin; 0.93 ± 0.32 vs. 0.77 ± 0.27 mg/kg/day, P < 0.001; alogliptin; 0.37 ± 0.15 vs. 0.30 ± 0.13 mg/kg/day, P = 0.002). The associations between adverse symptoms and the higher daily doses of sitagliptin (the most common DPP-4 inhibitor used in this study) were not observed among all subjects, male or female subjects, or subjects with liver disease (Supplemental Table 6).

4 Discussion

In this study, mild hypoglycemic symptoms were reported in 34 (2.2 %) of the total subjects under sitagliptin or alogliptin therapy, among the four DPP-4 inhibitor therapy groups. The risk factors associated with hypoglycemic symptoms include liver disease, female sex and alcohol consumption more than three times per week. Interestingly, all of these factors have been identified as physiological or pathological predisposing factors for hypoglycemia [8, 12–16]. No hypoglycemic symptoms were reported under vildagliptin or linagliptin therapy, which might be due to the small number of patients who were prescribed these drugs in this study. The lower prescription rates (linagliptin: 0.3 %, vildagliptin: 8.8 %) compared with sitagliptin (75.4 %) might result from the differences in the dates when these drugs were marketed in Japan, i.e. September 2011 for linagliptin, April 2010 for vildagliptin and December 2009 for sitagliptin.

DM and nonalcoholic fatty liver disease (NAFLD) appear to have common origins related to obesity and insulin resistance, and DM is also common among patients with alcoholic and viral chronic liver disease (CLD). Abnormalities of glucagon secretion are often observed in patients with DM or CLD and may actually reflect an impairment of α-cell glucose sensing [17–19]. Glucagon secretion plays an essential role in the regulation of hepatic glucose production and in the counter-regulatory response to hypoglycemia [17, 18]. Consequently, patients with both DM and NAFLD or CLD are at high risk for hypoglycemia regardless of the severity of liver disease [8, 15, 16]. Sulfonylurea and insulin dose adjustment is required in patients with DM and CLD, whereas DPP-4 inhibitor dose adjustment is not [1, 15]. Evidence exists to suggest that DPP-4 is an important player in the pathogenesis of NAFLD, and DPP-4 inhibitors therefore have considerable therapeutic potential for CLD [19, 20]. However, our results suggest that patients with DM and NAFLD or CLD being treated with DPP-4 inhibitor therapy must be counseled carefully regarding recognition and management of hypoglycemia in order to maximize efficacy.

This is the first report to suggest that females are more susceptible to DPP-4-induced hypoglycemia. Elucidation of sex differences in the physiology and pathophysiology of glucose homeostasis is a key issue for gender-specific care of DM [21]. In the past 30 years, the all-cause mortality and cardiovascular mortality rates for females with DM have not declined, in contrast to the decreased rates observed in males [21]. Females have an increased risk of experiencing hypoglycemic events requiring medical intervention in the emergency department or outpatient setting [8]. Reduced counter-regulatory hormonal responses to hypoglycemia in females have been postulated to be a risk factor [8, 22]. Physiologically, females have lower fasting plasma glucose levels [21, 23]. Lower rates of gluconeogenesis in females in the exercising and/or fasting state may be the result of lower increments in glucagon secretion in females than in males [22]. On the other hand, we speculated that no associations between hypoglycemic symptoms and the dose of the therapeutic agents were found due to the specific mechanism of DPP-4 inhibitors, i.e., that DPP-4 inhibitors may improve the ability of both α- and β-cells to sense and respond appropriately to hypoglycemia [24]. Nevertheless, further investigations are needed to confirm whether the dose being used for the treatment is associated with hypoglycemia.

Alcohol intake is known to be a risk factor for hypoglycemia, particularly in patients who take oral glucose-lowering agents [12, 13, 25]. Ethanol delays recovery from hypoglycemia by impeding counter-regulatory hormonal responses and enhances glucose-lowering effects by decreasing hepatic gluconeogenesis [16, 25]. The effects of chronic ethanol consumption may be pronounced in females [14]. In the present study, however, neither alcohol intake nor female sex was related to hypoglycemic symptoms among the patients receiving monotherapy. These results suggest that the effects of alcohol intake and female sex become apparent only in subjects taking other glucose-lowering agents.

In this study, nonhypoglycemic symptoms were found to be associated with combination therapy. Age was not found to be a risk factor for DPP-4-related hypoglycemic and nonhypoglycemic symptoms, even though most of the subjects were over 65 years of age (mean age: 68.0 ± 11.8 years, range 19–98 years) and did not receive dose adjustments. The risk of hypoglycemia in older subjects is higher because of the physiology of aging and cognitive decline [12, 26]. The risk might be at least partly masked in this study because self-reports can be affected by lower perceptions of hypoglycemia among the elderly. Meanwhile, there is much clinical evidence supporting the efficacy and safety of DDP-4 inhibitor therapy in older patients [26].

This investigation has some obvious limitations. First, a cross-sectional design does not allow causality between the drugs and the ADRs to be established. The ADRs were reported largely on the basis of the patient’s memory; therefore, they may have been overestimated. It is important to consider the “nocebo” effect, that is, a patient’s negative expectations of experiencing adverse effects [27]. Second, we did not have adequate information regarding complications, laboratory evidence or medication history. The prevalence of chronic liver disease (4.1 %) in this study was obviously lower than that reported in DM patients [15]. This is because the presence of liver disease was judged according to the prescription of antiviral agents and/or hepatoprotective agents, such as ursodeoxycholic acid. These findings are therefore biased based on the proxy assessment with high specificity and low sensitivity. Third, in this study, we defined ‘hypoglycemic symptoms’ on a theoretical basis, according to a previous study [12]. Moreover, we defined a single symptom as a ‘hypoglycemic symptom’ in most cases since few participants had more than one hypoglycemic symptom. Therefore, we could not exclude the possibility that there were misclassifications of symptoms in some participants. In addition, we excluded 452 patients because of a lack of data; therefore, a selection bias in the study population may exist, although the patient characteristics were not significantly different between the included and excluded patients. Lastly, the follow-up period for investigating ADRs was short (i.e., the median interval was 28 days), which might have contributed to the low incidence of ADRs observed in this study. In this study, the adjusted regression models explained only a small part of the variance of these risks (Nagelkerke’s R squared: 2.0–9.4 %), because all potential risk factors could not be included and the incidence of ADRs was relatively small. Therefore, the results of this study should be verified by clinical studies of longer duration, more parameters and with a larger sample size.

5 Conclusion

This study suggests that mild hypoglycemic symptoms, as reported by 2.2 % of the total subjects treated with sitagliptin or alogliptin therapy, are associated with liver disease in patients receiving monotherapy, and female sex and alcohol consumption in patients receiving combination therapy. All of these factors are potentially capable of causing poor gluconeogenesis as a result of decreased counter-regulatory hormonal responses to hypoglycemia. Health care providers should thus monitor hypoglycemic symptoms carefully in patients with these potential risk factors, and clinical studies of higher quality and with a larger sample size should be undertaken to confirm the present findings.