Reduced vascular events in type 2 diabetes by biguanide relative to sulfonylurea: study in a Japanese Hospital Database

The study protocol was approved by the Review Board on Clinical Research of Fukuoka University (Fukuoka, Japan). This retrospective observational study was done using a hospital-based composite database stored in hospital electronic-information systems constructed by MDV. MDV data were purchased by FUJIFILM Pharma Co., Ltd. A contract between FUJIFILM Pharma Co., Ltd., and Fukuoka University Hospital enabled analyses to be carried out. The current database of MDV has been expanded to a much greater extent than that reported previously [15]. The database of the medical information of patients (including laboratory data) was extracted from the medical-cost accounting system of 103 institutions (19 hospitals with <200 beds; 66 hospitals with 200–499 beds; and 18 hospitals with ≥500 beds) in Japan. This database contains information about age, sex, diagnosis, International Classification of Diseases (ICD)-10 code, surgical history, outpatient/inpatient status, prescriptions, and laboratory data. Written informed consent was not obtained from each patient because all data were extracted from the database retrospectively. Nevertheless, patient anonymity was guaranteed.

Two studies (study 1 and study 2) were undertaken. The population of patients extracted in the present study satisfied the following conditions: (i) began treatment with a single OHA from 1 April 2008 to 30 April 2013; (ii) HbA1c level (NGSP) at baseline was available; (iii) age at baseline was 40–70 years; (iv) the presence or absence of CVD history was not considered in study 1, but the presence of CVD history was considered in study 2.

Exclusion criteria were: (i) began treatment with insulin, GLP-1 analog, or compounding agent at baseline; (ii) began treatment with ≥2 types of OHA; (iii) admitted to hospital at baseline; (iv) an anti-cancer drug was used before baseline measurements were taken; (v) The patient who was judged having a missing data by MDV Co., Ltd. The MDV database in 2013 contained the data of 225,197 individuals. Among them, the baseline value of HbA1c was available for 29,074 patients. After careful adherence to inclusion and exclusion criteria, 4095 and 1273 individuals were subjected to cohort analyses of study 1 and study 2, respectively. Outlines as well as the classification of inclusion and exclusion in study 1 and study 2 are summarized in Figs. 1 and 2. Patients with cancer or liver cirrhosis were not listed in any drug groups. However, renal failure was noted in study 1: 2/629 (0.3 %) in the SU group, 4/1305 (0.3 %) in the BG group, 2/592 (0.3 %) in the α-GI group, 2/351 (0.6 %) in the TZD group, 2/223 (0.9 %) in the glinide group, and 6/995 (0.6 %) in the DPP-4 inhibitor group. Renal failure was also noted in study 2: 1/228 (0.4 %) in the SU group, 0/325 (0 %) in the BG group, 0/203 (0 %) in the α-GI group, 1/140 (0.7 %) in the TZD group, 2/70 (2.9 %) in the glinide group, and 6/307 (2.0 %) in the DPP-4 inhibitor group. However, there was no significant difference in the involvement of renal failure among the above 6 groups by Chi-squared test, suggesting little impact on the analysis. So we retained these data for the analysis.

Information regarding cardiovascular events was obtained from the MDV database. A cardiovascular event was defined as suffering angina pectoris (codes I200, I201, I208 and I209), MI (I210–I213, I219, I220, I221 and I229), heart failure (I500, I501 and I509), cerebral infarction (I630–I635, I638 and I639), cerebral hemorrhage (I614 and I619), or subarachnoid hemorrhage (I600–I602, I604, I605 and I609).

Observation was stopped at the time of: (i) a cardiovascular event; (ii) changing of the OHA at baseline; (iii) discontinuation of the drug at baseline before a prescription period of 2 years. However, observation was not suspended if a new OHA was added to the first drug. Finally, 1246 subjects and 412 cases in study 1 and study 2, respectively, were treated continuously by a baseline drug without a cardiovascular event being recorded for >2 years (summarized in Figs. 1 and 2).

Baseline characteristics of patients (sex, age, HbA1c level, use of other drugs (anti-hypertension or anti-dyslipidemia)) and cardiovascular events were analyzed. Subjects were divided into six groups according to the baseline drug. Prevalence of CVD and change in HbA1c level from baseline in each drug group was assessed.

To verify the independence of the analysis in the present study, statistical evaluation was undertaken by a specialist organization (ING Corp., Tokyo, Japan). ING Corp., are in no way associated with our institution or any drug company (including FUJIFILM Pharma Co., Ltd.). Statistical analyses were done using SAS v9.1.3 for Windows (SAS Institute, Cary, NC, USA). Prevalence of CVD in each group was investigated by Kaplan-Meier analyses. Comparisons between groups were achieved using the Cox-proportional hazard model adjusted for sex, age, HbA1c level (NGSP), use or non-use of anti-hypertensive drugs, and use or non-use of anti-dyslipidemia drugs. For consideration of the type-1 error in the Student’s t-test, patient characteristics at baseline between two groups (SU vs BG, SU vs α-GI, SU vs TZD, SU vs glinide, and SU vs DPP-4 inhibitor) were compared by the Dunnett test. P < 0.05 was considered significant.

The Japanese Patient Survey of 2011 for patients with DM was published by the MHLWJ. Hence, we compared our extracted database from 2013 with that of the MHLWJ. Our database contained 225,197 subjects and the percentage of males was 60.7 %. Based on data from the MHLWJ, the estimated number of subjects with DM in Japan was 2,700,000, and the percentage of males was 55.1 %. Age distribution of subjects with DM in our database was 5.8 % in their forties, 13.8 % in their fifties, 28.6 % in their sixties, 29.9 % in their seventies, and 15.8 % in their eighties, whereas that for the MHLWJ was 5.6, 13.7, 32.4, 31.1, and 13.4 %, respectively. The percentage of males in our database was 71.5 % in their forties, 68.7 % in their fifties, 65.6 % in their sixties, 60.2 % in their seventies, and 47.3 % in their eighties, whereas that of MHLWJ was 68.2, 61.2, 57.8, 53.8 and 40.9 %, respectively. These results suggested that our database could reflect the current situation of DM in Japan, as reported previously.

Baseline characteristics of 4095 subjects in study 1 and 1273 subjects in study 2, including the six groups based on the type of OHA (i.e., SU, biguanide, α-GI, TZD, glinide, DPP-4 inhibitor) are shown in Tables 1 and 2, respectively. Mean age of biguanide and TZD groups was slightly (but significantly) younger than the SU group in study 1 and study 2. Treatment with anti-hypertension and anti-dyslipidemia drugs was undertaken in all OHA groups. The detailed classification of antihypertensive drugs and anti-dyslipidemia drugs used in subjects of study 1 and 2 are shown in Additional files 1 and 2, respectively. Mean HbA1c levels (NGSP) of biguanide, α-GI, TZD and glinide were slightly (but significantly) lower than those of the SU group in study 1 and study 2. Prevalence of patients with a high HbA1c value (≥6.5 %) was not significantly different between the SU group and the other groups in study 1 and study 2. Mean duration of administration of baseline drugs was 320–457 days in study 1 and 330–540 days in study 2. In study 1, the duration of administration of BG, α-GI, TZD and glinide was not significantly different from that of SU, but that of DPP-4 inhibitor was significantly shorter than that of SU. In study 2, the duration of administration of BG was significantly longer than that of SU, and that of DPP-4 inhibitor was significantly shorter than that of SU.

Detailed data on the prevalence of cardiovascular events in study 1 and study 2 are shown in Table 3A and B, respectively. Based on these data, composite (cardiac and cerebrovascular) cardiovascular events were used mainly for statistical analyses subsequently because meaningful values could not be obtained in the analysis of cerebral events only due to the small number of events. The number of cardiovascular events (1000 person-years) is shown in Table 4A and B.

Kaplan-Meier curves for T2DM patients who experienced ≥1 cardiovascular event in each group and statistical comparisons between SU and other drugs in study 1 are shown in Fig. 3. For determination of the hazard ratio (HR), the Cox proportional hazard model was adjusted for sex, age, HbA1c value (NGSP), use or non-use of anti-hypertensive drugs, and use or non-use of anti-dyslipidemia drugs. During a 104-week observation period, patients who began treatment with BG showed a significantly lower risk of cardiovascular events compared with those who started treatment with SU. HR of BG relative to SU was 0.603 (95 % confidence interval [CI], 0.439–0.829, P = 0.002). Significant risk reduction by BG relative to SU was also observed in analyses of cardiac events only (HR, 0.613; 95 % CI, 0.438–0.857; P = 0.004). However, α-GI, TZD, glinide and DPP-4 inhibitor did not show such significant reduction in risk compared with SU in terms of cardiovascular events and cardiac events (data not shown). A similar analysis using Kaplan-Meier curves targeting T2DM patients who did not have a history of cardiovascular events (n = 2822) did not show a significant difference in the prevalence of cardiovascular events between SU and other OHAs (data not shown).

Changes in HbA1c levels from baseline (0 week) to 2 years (104 weeks) in patients in whom a cardiovascular event was observed were compared among OHAs. In patients who began treatment with SU, BG, α-GI, a TZD, glinide or DPP-4 inhibitor, the HbA1c level was reduced with SU, BG, α-GI, a TZD and DPP-4 inhibitor, but was increased with glinide. The maximum lowering effect in HbA1c level was observed in the DPP-4 inhibitor group. The difference in the change in HbA1c level between SU (−0.31 ± 0.07, N = 236) and glinide (0.08 ± 0.12, N = 81), and between SU and DPP-4 inhibitor (−0.75 ± 0.11, N = 95) was significant (P = 0.0005 and P < 0.0001, respectively). However, the extent of change in the level of BG (−0.30 ± 0.05, N = 542), α-GI (−0.21 ± 0.08, N = 195) and TZD (−0.31 ± 0.11, N = 92) relative to that of SU was not significant (P = 0.90, P = 0.37 and P = 0.95, respectively). Similar results were observed in study 2. The difference in the change in HbA1c level between SU (−0.43 ± 0.11, N = 80) and TZD (−0.003 ± 0.16, N = 42) and between SU and glinide (0.50 ± 0.20, N = 24) was significant (P = 0.029 and P < 0.001, respectively). However, the extent of change in the level of BG (−0.19 ± 0.08, N = 159), α-GI (−0.19 ± 0.11, N = 80) and DPP-4 inhibitor (−0.27 ± 0.20, N = 27) relative to that of SU was not significant (P = 0.069, P = 0.12 and P = 0.48, respectively).