Diabetes Ther (2018) 9:1477–1490 https://doi.org/10.1007/s13300-018-0452-y ORIGINAL RESEARCH Comparative Effectiveness of DPP-4 Inhibitors Versus Sulfonylurea for the Treatment of Type 2 Diabetes in Routine Clinical Practice: A Retrospective Multicenter Real-World Study . . . Gian Paolo Fadini Daniele Bottigliengo Federica D’Angelo . . . Franco Cavalot Antonio Carlo Bossi Giancarlo Zatti . . Ileana Baldi Angelo Avogaro On behalf of the DARWIN-T2D Network Received: May 3, 2018 / Published online: June 1, 2018 The Author(s) 2018 Methods: This was a multicenter retrospective ABSTRACT study on diabetes outpatient clinics comparing the effectiveness of DPP4i versus gliclazide Introduction: DPP-4 inhibitors (DPP4i) and extended release. The primary endpoint was sulfonylureas are popular second-line therapies change from baseline in HbA1c. Secondary for type 2 diabetes (T2D), but there is a paucity endpoints were changes in fasting plasma glu- of real-world studies comparing their effective- cose, body weight, and systolic blood pressure. ness in routine clinical practice. Automated software extracted data from the same clinical electronic chart system at all centers. Propensity score matching (PSM) was Enhanced digital features To view enhanced digital features for this article go to https://doi.org/10.6084/ used to generate comparable cohorts to perform m9.ﬁgshare.6292265. outcome analysis. Results: We included data on 2410 patients See ‘‘Acknowledgements’’ for members of the DARWIN- starting DPP4i and 1590 patients starting gli- T2D Network. clazide (mainly 30–60 mg/day). At baseline, the Electronic supplementary material The online two groups differed in disease duration, body version of this article (https://doi.org/10.1007/s13300- weight, blood pressure, HbA1c, fasting glucose, 018-0452-y) contains supplementary material, which is HDL cholesterol, triglycerides, liver enzymes, available to authorized users. eGFR, prevalence of microangiopathy, and use of metformin. Among DPP4i molecules, no G. P. Fadini (&) G. Zatti A. Avogaro Department of Medicine, University of Padova, difference in glycemic effectiveness was detec- 35128 Padua, Italy ted. In matched cohorts (n = 1316/group), e-mail: firstname.lastname@example.org; patients starting DPP4i, as compared with email@example.com patients starting gliclazide, experienced greater D. Bottigliengo I. Baldi reductions in HbA1c (- 0.6% versus - 0.4%; Department of Cardiac Thoracic and Vascular p \ 0.001), fasting glucose (- 14.1 mg/dl versus Sciences, University of Padova, 35128 Padua, Italy - 8.8 mg/dl; p = 0.007), and body weight F. D’Angelo (- 0.4 kg versus - 0.1 kg; p = 0.006) after an INRCA IRCSS Ancona, 60124 Ancona, Italy average 6 months follow-up. DPP4i improved glucose control more than gliclazide, especially F. Cavalot San Luigi Gonzaga Hospital, 10043 Orbassano, Italy in patients who had failed with other glucose- lowering medications or were on basal insulin. A. C. Bossi ASST Bergamo Ovest, 24047 Treviglio, Italy 1478 Diabetes Ther (2018) 9:1477–1490 Conclusions: This large retrospective real-world Furthermore, gliclazide is by far the mostly study shows that, in routine clinical practice, widely used SU in Italy  and its relative starting a DPP4i allows better glycemic control cardiovascular safety has been recently con- than starting low-dose gliclazide. ﬁrmed in the Italian TOSCA.IT study . Funding: The Italian Diabetes Society, with Despite SUs and DPP4i being very popular external support from AstraZeneca. GLMs, there is a striking paucity of real-world studies comparing the effectiveness of these drugs in routine clinical practice. This is par- Keywords: Database; Epidemiology; ticularly important since SUs are still perceived Pharmacotherapy as highly effective drugs for the control of hyperglycemia and their low cost makes them INTRODUCTION particularly attractive for healthcare systems with limited resources. Thus, comparative real- The rapid evolution of diabetes pharmacother- world studies on these drugs may complement apy and the availability of many glucose-low- information from RCTs and inform on thera- ering medication (GLM) classes for the peutic appropriateness. The DARWIN-T2D was a multicenter retro- treatment of type 2 diabetes (T2D) have made the choice of second-line agents after met- spective study conducted on electronic medical records containing clinical data, performed at formin an difﬁcult task . Ideally, patient characteristics should be matched with drug 46 diabetes specialist outpatient clinics in Italy modes of action, favorable effects, and side . We herein report results of a subanalysis effects. However, this is not always possible comparing the effectiveness of DPP4i versus because of restrictions in drug indications, gliclazide on glycemic and extra-glycemic availability, reimbursement, and contraindica- endpoints. tions. Furthermore, translating results of phase III randomized controlled trials (RCTs) to clini- METHODS cal practice may be problematic owing to the many differences between trial and routine care Data Source settings . In many countries, the two most popular The main objective of the DARWIN-T2D study second-line GLMs for the treatment of T2D are sulfonylureas (SUs) and dipeptidyl peptidase 4 was to describe the clinical characteristics and the changes from baseline in glycemic and inhibitors (DPP4i). RCTs that directly compared drugs of these classes showed that SUs tend to extra-glycemic effectiveness parameters in patients newly treated with the SGLT2 inhibitor be more effective than DPP4i in reducing HbA1c over the short term, but that such differences dapagliﬂozin, a DPP4i, gliclazide, or a GLP-1 receptor agonist . The study was conducted are mostly lost in the long run . Rather, at 46 Italian diabetes outpatient clinics. The compared with SU, DPP4i have been associated detailed study protocol and the primary results with a markedly lower risk of hypoglycemia and have been previously published . Results of a mild beneﬁt in the control of body weight . the study indicated a signiﬁcant channelling of In addition, while three large placebo-con- different patients towards different GLMs, and trolled RCTs support the cardiovascular safety of DPP4i [5–7], SUs have been linked with an an overall low common support between patients receiving dapagliﬂozin and other GLMs increased risk of cardiovascular events and mortality, although such data derive mainly . The largest common support of propensity scores was detected for patients starting DPP4i from observational studies [8, 9]. Gliclazide is and patients starting gliclazide, thereby pro- often considered the preferred SU, because it viding a rationale for comparing effectiveness of has been associated with greater than 50% less such drugs. hypoglycemia risk  and a safer cardiovascu- lar risk proﬁle  compared with other SUs. Diabetes Ther (2018) 9:1477–1490 1479 Patients were retrospectively included if they electronic chart system (MyStar Connect, were aged 18–80 years, had a diagnosis of T2D Me.Te.Da.). for at least 1 year, and were newly prescribed with a full-dose DPP4i (per protocol, linagliptin Multiple Imputation and Propensity Score was excluded ), or with gliclazide extended Matching release at a daily dose of 30 mg or higher. Exclusion criteria were a diagnosis of type 1 For a comparative analysis of effectiveness, we diabetes and age less than 18 or greater than used propensity score matching (PSM), one of 80 years. the most popular methods to estimate treat- Dedicated software automatically extracted ment effects in observational studies . In a all relevant clinical data (demographics, trade-off between unconfoundedness and pre- anthropometrics, blood pressure, HbA1c, fast- cision, the following baseline covariates were ing plasma glucose, lipid values, liver enzymes, chosen for PSM as they are expected to affect renal function, history of complications, and outcomes and therapy assignment: age, gender, medications) at baseline and at the ﬁrst avail- diabetes duration, BMI, body weight, systolic able follow-up visit, 3–12 months after baseline. and diastolic blood pressure, FPG, Hb1Ac, total LDL cholesterol levels were calculated using cholesterol, HDL cholesterol, triglycerides, Friedewald’s equation , whereas eGFR (esti- aspartate aminotransferase, alanine amino- mated glomerular ﬁltration rate) was computed transferase, eGFR, insulin as associated therapy, using the CKD-EPI equation . Microan- metformin as associated therapy, use of DPP4i giopathy was deﬁned as the presence of an or gliclazide as second-line therapy, microan- albumin excretion rate greater than 30 mg/24 h giopathy and macroangiopathy. Presence of or mg/g of creatinine, an eGFR less than 60 ml/ missing data was handled with multiple impu- min/1.73 m , diabetic neuropathy (either tation (MI), as previously described . Out- somatic or autonomic), diabetic retinopathy comes and selected variables were used as (any grade), or maculopathy. Macroangiopathy predictors in MI models . A PS model was was deﬁned as the presence of a history of ﬁtted on each imputed data set and the ﬁnal myocardial infarction or stroke/transient individual PS value was computed as the aver- ischemic attack, peripheral arterial disease, sur- age of all the subject PS values obtained in each gical or endovascular revascularization (any imputed data set. Then, PS values were used to site), or a diagnosis of asymptomatic create a matched set of individuals from the atherosclerosis. We retrieved information on all original non-imputed data set. Matching was concomitant medications and on the entire performed with 1:1 ratio, i.e., each subject history of GLM use to deﬁne whether patients treated with DPP4i was matched with only one were being prescribed DPP4i or gliclazide as subject treated with gliclazide, using a genetic second-line agents after metformin (i.e., had algorithm, without replacement. Covariate bal- been treated only with metformin) or after ance after matching was evaluated using stan- failure of at least another GLM different from dardized mean difference across group of metformin. treatment and standardized mean differences of The primary effectiveness endpoint was the the square of continuous variables. Balance was change from baseline in HbA1c. Secondary achieved if standardized difference was less than endpoints were changes from baseline in fasting 0.1. Outcome analysis was conducted on a plasma glucose, body weight, and systolic blood matched set of individuals obtained after PSM. pressure. We excluded patients without a fol- Effect of treatment on outcomes was evaluated low-up examination, those with missing data with adjusted linear regression models, with for the primary outcome at baseline or follow- conﬁdence intervals computed using a robust up, and those initiating DPP4i and gliclazide at sandwich estimator. More details on MI and the same visit (because the effect could not be PSM can be found in the Online Supplementary attributed to one or the other). All data were Material. extracted automatically from the same 1480 Diabetes Ther (2018) 9:1477–1490 Statistical Analysis baseline data on 6594 T2D patients who initiated therapy with a DPP4i (53.2% sitagliptin, 22.9% alogliptin, 20.6% vildagliptin, 3.3% saxagliptin) Except where otherwise speciﬁed, data are pre- and 5960 patients who initiated therapy with sented as mean ± standard deviation or as per- gliclazide. Of these, 2999 patients treated with centage, as appropriate. Comparisons between DPP4i (45.5%) and 2111 patients treated with the two groups of patients (e.g., those receiving gliclazide (35.4%), had a follow-up visit available DPP4i and those receiving gliclazide) were per- between 3 and 12 months after baseline; 589 formed using the 2-tail unpaired Student t test for patients treated with DPP4i and 521 treated with continuous variables, or the chi-square test for gliclazide were excluded for missing data for the categorical variables. Differences in clinical char- primary outcome or because they initiated both acteristics between matched cohorts were better drugs at the same time (n = 151). Data on the analyzed using standardized bias than using p remaining 2410 DPP4i users and 1590 gliclazide values, as previously suggested . Comparisons users are shown in Table 1. Patients newly treated in continuous variables among more than two with gliclazide versus those newly treated with groups was performed using ANOVA with Bon- DPP4i had longer disease duration, higher body ferroni correction. Evaluation of within-group weight, BMI, systolic blood pressure, HbA1c, changes in outcome variables was performed fasting plasma glucose, triglycerides, and liver using the 2-tail paired Student’s t test. Changes enzymes, and lower HDL cholesterol and eGFR. from baseline in outcome variables were calcu- Patients starting gliclazide also had a higher lated for each group as data collected at follow-up prevalence of microangiopathy and less frequent minus data collected at baseline, and compared use of metformin than patients starting a DPP4i. using the 2-tail unpaired Student t test. To analyze the time trend of HbA1c reduction in the two groups, we divided the 9-month observation Within-Group Effectiveness Analysis window (3–12 months after baseline) into ﬁve equal periods and assigned each patient to the After a median follow-up of 6.1 months (IQR relevant follow-up duration. Statistical signiﬁ- 5.5–6.7), in patients who received a DPP4i, cance was accepted at p\ 0.05. HbA1c declined by 0.6%, fasting plasma glucose declined by 11.4 mg/dl, body weight declined Compliance with Ethics Guidelines by 0.5 kg, with no signiﬁcant change in systolic blood pressure (Table S1). Among DPP4i, no signiﬁcant difference was observed in the The study was approved by the ethical committee change from baseline in HbA1c, fasting plasma of each participating center. All procedures per- glucose, and systolic blood pressure, while formed in studies involving human participants reductions in body weight were larger for sita- were in accordance with the ethical standards of gliptin and alogliptin than for vildagliptin the institutional and/or national research com- (Fig. S1). mittee and with the 1964 Declaration of Helsinki After a median follow-up of 6.2 months (IQR and its later amendments or comparable ethical 4.8–7.1), in patients who received gliclazide, standards. Since the study was performed retro- HbA1c declined by 0.6% and fasting plasma spectively on an anonymized database, no patient glucose declined by 14.5 mg/dl, while no sig- consent was required. niﬁcant change was observed for body weight and blood pressure (Table S1). RESULTS Comparison of Propensity Score Matched Study Population Groups The study ﬂowchart is shown in Fig. 1. Between 15 PSM was performed on a predeﬁned set of March 2015 and 31 December 2016, we collected variables, which were considered to be clinically Diabetes Ther (2018) 9:1477–1490 1481 Fig. 1 Study ﬂowchart. DPP4i dipeptidyl peptidase 4 inhibitors, PSM propensity score matching relevant for the outcome and therapy assign- lower glycemic effect with longer follow-up, ment. After MI, PSM identiﬁed 1316 patients in such loss of effectiveness was not observed with each group, who were well balanced for all DPP4i (Fig. S4). clinical variables except concomitant use of After we adjusted for concomitant insulin basal insulin, which was signiﬁcantly more use, patients starting DPP4i still showed greater common in patients starting a DPP4i (17.4%) reductions in HbA1c and body weight than versus those starting gliclazide (13.2%) (Fig. S2). those starting gliclazide (Fig. S5). In models In matched cohorts (Fig. 2), the change from fully adjusted for basal insulin, each variable at baseline in HbA1c was signiﬁcantly higher in baseline (either linearly or non-linearly mod- patients starting DPP4i than in those starting elled), and interaction terms, DPP4i proved gliclazide (- 0.6 ± 1.1% versus - 0.4 ± 1.2%; superior to gliclazide in reducing HbA1c, FPG, p \ 0.001). The same was true for fasting plasma body weight, and systolic blood pressure glucose (- 14.1 ± 43.5 mg/dl versus (Table S2). - 8.8 ± 46.2 mg/dl; p = 0.007) and body weight Figure 3 shows the changes in HbA1c, FPG, (- 0.4 ± 3.3 kg versus - 0.1 ± 2.9 kg; body weight, and systolic blood pressure in the p = 0.006), while the between-group difference matched cohorts of patients with or without in the change from baseline in systolic blood concomitant insulin therapy and in those ini- pressure did not reach statistical signiﬁcance tiating DPP4i or gliclazide as second-line ther- (- 1.5 ± 19.8 mmHg versus 0.3 ± 18.9 mmHg; apy after metformin or as a more advanced line p = 0.056). of therapy. Patients starting DPP4i experienced The use of concomitant GLM did not sig- a stronger HbA1c reduction than patients niﬁcantly change at follow-up compared to starting gliclazide, irrespective of background baseline in either group (Fig. S3). insulin therapy, but the between-group differ- By dividing patients according to distance ence was signiﬁcantly larger in insulin-treated between baseline and follow-up, we simulated a patients. Only in patients who were on insulin time course of HbA1c reduction: while patients therapy, DPP4i reduced fasting plasma glucose who received gliclazide show a progressively more than gliclazide. Vice versa, only in 1482 Diabetes Ther (2018) 9:1477–1490 Table 1 Baseline clinical characteristics of study patients Before PSM After PSM DPP4i Gliclazide SB p DPP4i Gliclazide SB p Number 2410 1590 – 1316 1316 – Age, years 67.0 ± 9.2 67.4 ± 9.0 0.051 0.116 67.6 ± 9.2 67.7 ± 8.9 0.012 0.760 Sex male, % 59.5 59.3 0.003 0.924 59.6 59.8 0.005 0.905 Current smoker, % 18.5 19.4 0.023 0.535 19.5 18.1 0.036 0.676 Diabetes duration, years 11.0 ± 7.6 11.7 ± 7.4 0.096 0.003 11.8 ± 7.9 11.9 ± 7.6 0.015 0.709 Weight, kg 79.9 ± 15.1 82.3 ± 15.9 0.158 \ 0.001 80.5 ± 15.5 81.3 ± 15.0 0.053 0.193 Height, cm 165.6 ± 9.5 166.0 ± 9.5 0.042 0.230 165.6 ± 9.8 166.1 ± 9.4 0.052 0.237 BMI, kg/m 29.1 ± 4.9 29.8 ± 5.0 0.143 \ 0.001 29.3 ± 4.9 29.5 ± 4.8 0.029 0.479 Waist circumference, cm 103.1 ± 12.2 104.2 ± 12.0 0.091 0.086 103.5 ± 11.7 103.7 ± 11.7 0.017 0.730 SBP, mmHg 136.7 ± 17.8 139.6 ± 19.7 0.152 \ 0.001 138.4 ± 18.5 139.4 ± 19.5 0.049 0.275 DBP, mmHg 78.6 ± 9.0 78.9 ± 9.3 0.026 0.478 79.1 ± 9.4 78.5 ± 9.2 0.063 0.165 FPG, mg/dl 152.8 ± 36.0 166.4 ± 43.9 0.338 \ 0.001 159.0 ± 40.8 158.2 ± 39.2 0.020 0.632 HbA1c, % 7.8 ± 0.9 8.1 ± 1.2 0.358 \ 0.001 7.9 ± 1.0 7.9 ± 1.0 0.002 0.955 Total cholesterol, mg/dl 170.5 ± 35.2 172.9 ± 38.4 0.063 0.095 171.7 ± 36.0 169.6 ± 36.3 0.058 0.214 HDL cholesterol, mg/dl 48.8 ± 13.6 47.5 ± 13.3 0.094 0.017 48.9 ± 14.0 48.1 ± 13.3 0.055 0.241 Triglycerides, mg/dl 140.1 ± 80.9 150.4 ± 97.1 0.116 0.002 145.4 ± 88.6 142.6 ± 77.7 0.035 0.460 LDL cholesterol, mg/dl 93.9 ± 30.2 95.6 ± 32.6 0.054 0.146 94.1 ± 30.4 93.4 ± 31.7 0.022 0.668 SGOT, U/l 21.9 ± 10.9 23.2 ± 11.8 0.116 0.019 21.7 ± 10.8 22.8 ± 11.0 0.103 0.083 SGPT, U/l 26.3 ± 15.8 28.2 ± 18.9 0.106 0.028 26.0 ± 15.4 27.5 ± 18.2 0.092 0.122 Creatinine, mg/dl 0.9 ± 0.3 0.9 ± 0.3 0.000 0.253 0.9 ± 0.3 0.9 ± 0.3 0.000 0.276 eGFR, mg/min/1.73 m 81.0 ± 18.3 78.6 ± 18.4 0.135 0.001 78.4 ± 19.4 77.9 ± 18.5 0.029 0.597 AER, mg/24 h 64.2 ± 223.5 62.9 ± 205.4 0.006 0.900 77.9 ± 278.8 63.1 ± 213.0 0.060 0.298 Diabetes Ther (2018) 9:1477–1490 1483 Table 1 continued Before PSM After PSM DPP4i Gliclazide SB p DPP4i Gliclazide SB p Complications Nephropathy, % eGFR \ 60 ml/min/1.73 m 13.7 18.4 0.130 0.002 17.7 19.3 0.041 0.396 AER [ 30 mg/24 h 26.9 27.0 0.002 0.950 28.3 25.9 0.054 0.358 Retinopathy, % Any retinopathy 11.7 15.0 0.098 0.007 14.2 14.9 0.020 0.638 Macular edema 2.0 2.8 0.053 0.123 2.7 3.2 0.030 0.530 Neuropathy, % Somatic 15.5 21.1 0.147 0.006 17.7 20.2 0.064 0.350 Autonomic 2.3 1.3 0.073 0.159 2.4 1.2 0.090 0.157 Peripheral arterial disease, % Arteriosclerosis obliterans 15.9 13.9 0.056 0.322 18.2 14.7 0.095 0.162 Previous revascularization 2.4 1.2 0.087 0.110 3.1 1.3 0.123 0.046 Foot disease, % Active ulcers 2.2 2.0 0.014 0.823 3.0 1.9 0.071 0.214 Deformities 7.6 10.0 0.086 0.094 8.7 10.6 0.064 0.303 Previous amputation 1.1 0.9 0.020 0.685 1.5 0.9 0.055 0.382 Cerebrovascular disease, % Carotid atherosclerosis 38.7 43.7 0.102 0.019 41.9 44.6 0.055 0.281 Previous TIA/stroke 3.7 3.5 0.011 0.842 4.2 3.5 0.036 0.463 Coronary artery disease, % Ischemic heart disease 11.3 11.3 0.000 0.949 13.0 11.9 0.033 0.465 1484 Diabetes Ther (2018) 9:1477–1490 Table 1 continued Before PSM After PSM DPP4i Gliclazide SB p DPP4i Gliclazide SB p Previous revascularization 7.7 7.2 0.019 0.651 8.7 7.3 0.052 0.268 Microangiopathy, % 30.3 36.6 0.113 \ 0.001 35.2 36.5 \ 0.001 0.513 Macroangiopathy, % 36.2 39.4 0.066 0.056 39.9 40.6 0.014 0.743 Therapies Glucose lowering medications, % Metformin 85.6 76.6 0.236 \ 0.001 76.2 76.4 0.005 0.883 Basal insulin 13.9 14.3 0.010 0.550 17.4 13.2 0.118 0.002 Acarbose 1.4 1.4 0.000 0.926 1.8 1.4 0.032 0.400 Pioglitazone 5.6 6.3 0.030 0.309 6.3 5.7 0.025 0.546 Other therapy, % Antiplatelet 47.5 46.3 0.024 0.480 48.2 47.7 0.010 0.836 Statin 66.0 63.5 0.052 0.137 64.1 65.2 0.033 0.579 ACE/ARBs 65.0 67.8 0.059 0.101 67.5 67.5 0.000 0.983 CCB 20.9 23.1 0.053 0.139 21.9 22.9 0.024 0.598 Beta-blockers 28.5 30.1 0.035 0.307 29.9 30.4 0.011 0.816 Diuretics 10.7 13.4 0.083 0.022 12.3 13.4 0.033 0.466 Data are reported for the entire cohorts of patients starting dipeptidyl peptidase 4 inhibitors (DPP4i) or gliclazide and for the cohorts matched 1:1 after propensity score matching (PSM). Data are shown as mean ± SD or as percentage BMI body mass index, SBP systolic blood pressure, DBP diastolic blood pressure, FPG fasting plasma glucose, HDL high density lipoprotein, LDL low density lipoprotein, eGFR estimated glomerular ﬁltration rate, AER albumin excretion rate, TIA transient ischemic attack, ACEi angiotensin-converting enzyme inhibitors, ARB angiotensin receptor blockers, CCB calcium channel blockers, SB standardized bias, SGOT serum glutamic-oxaloacetic transaminase, SGPT serum glutamic pyruvic transaminase Diabetes Ther (2018) 9:1477–1490 1485 Fig. 2 Comparative effectiveness in matched cohorts. b), body weight (c), and systolic blood pressure (SBP, d). Baseline, follow-up data, and the change from baseline *p \ 0.05 for the indicated comparisons. Bars indicate are shown for the primary outcome (HbA1c, a) and for standard error secondary outcome measures: fasting plasma glucose (FPG, patients who were not on insulin, DPP4i ongoing therapy improved glucose control reduced body weight more than gliclazide. An more than addition of gliclazide. This was par- interaction between type of new prescription ticularly true in patients who were receiving (DPP4i vs gliclazide) and background insulin DPP4i or gliclazide after having failed with at therapy in determining the change from base- least another GLM different from metformin line in HbA1c, fasting plasma glucose, and body and in those who were on basal insulin. weight was conﬁrmed upon a multivariable These ﬁndings contrast with results of phase analysis (Table S2). III RCTs comparing DPP4i with SUs, which In patients who started DPP4i or gliclazide as show that DPP4i are less effective than SUs in second-line therapy after metformin, no signif- reducing HbA1c in the short term and are non- icant difference was noted in the changes from inferior to SUs in the long term. A few reasons baseline in HbA1c, fasting plasma glucose, body can explain these results. First, the broader weight, and systolic blood pressure. Rather, in population of patients included in this real- patients starting DPP4i or gliclazide as third or world study differs from that of phase III RCTs, more advanced line of therapy, reductions in especially in terms of age, history of GLM use, HbA1c, fasting plasma glucose, body weight, complication burden, and overall heterogene- and systolic blood pressure were greater with ity. Second, phase III RCT protocols require that DPP4i than with gliclazide. SUs are uptitrated to the maximal tolerated dose, which is rarely reached in clinical practice, especially in aged patients with chronic com- DISCUSSION plications. While we do not have full informa- tion about dose titration in the DARWIN-T2D This large retrospective real-world study study, gliclazide extended release is usually demonstrates that, in routine diabetes outpa- prescribed at the initial daily dose of 30–60 mg tient clinical practice, addition of a DPP4i to the 1486 Diabetes Ther (2018) 9:1477–1490 Fig. 3 Comparison of effectiveness according to concomi- presence or absence of concomitant basal insulin therapy, tant and previous therapy. The changes in HbA1c (a), and whether DPP4i or gliclazide was being used as second- fasting plasma glucose (FPG, b), body weight (c), and line therapy after metformin. *p \ 0.05 for the indicated systolic blood pressure (SBP, d) are shown for the entire comparisons. Bars indicate standard error cohorts of matched patients (all) or according to the and not uptitrated until the following visit. follow-up visits, we simulated a time course of Indeed, cross-sectional data from pilot study HbA1c change. Although this approach cannot centers indicate that the average daily gliclazide be equated to a longitudinal observation in the dose was 49 mg. Thus, it is possible that DPP4i same patients, it suggests that gliclazide lost allowed a better improvement in glucose con- effectiveness over time more than DPP4i, a trol because gliclazide was being used at a rela- trend present also in long-term phase III RCTs. tively low dose. As doses of gliclazide, but not of Remarkably, the history of previous glucose- DPP4i, can be uptitrated over time, it is possible lowering therapy had an impact on the com- that a longer observation would have produced parative effectiveness. In patients who were different results. It is nonetheless remarkable prescribed DPP4i or gliclazide as second-line that, at doses used in routine clinical practice, drugs after metformin, i.e., who had received no initiation of DPP4i provides a better glycemic GLM other than metformin, the improvement control than initiation of gliclazide at the ﬁrst in glucose control was similar with the two follow-up visit after about 6 months. treatment regimens. This is more in line with Of note, even in the larger cohort of patients results of phase III RCTs, wherein patients with before PSM, the various DPP4i provided similar T2D, usually of short duration, uncontrolled on improvements in HbA1c and fasting glucose metformin monotherapy, were randomized to and differed only in the change in body weight. DPP4i or SUs. However, in patients in whom at These data indicate that, from a practical per- least another GLM had failed, DPP4i was supe- spective, DPP4i do not differ in their glycemic rior to gliclazide in improving HbA1c, FPG, effectiveness. body weight, and blood pressure. Furthermore, Interestingly, by dividing study patients DPP4i retained a signiﬁcant glucose-lowering according to the time elapsed from baseline to effect when added to combination of basal Diabetes Ther (2018) 9:1477–1490 1487 insulin and oral therapy, which was greater non-users. Importantly, however, for as good as than the effect of gliclazide. The previous his- a PSM can be, it does not guarantee equal dis- tory of GLM use may be more reﬂective of tribution of unmeasured variables, making the residual beta-cell function than disease dura- issue of residual confounders unresolved. Data tion. Although both DPP4i and gliclazide stim- missingness was addressed with MI, but we ulate endogenous insulin secretion, DPP4i exert decided not to impute missing outcome vari- a more physiological meal-dependent action ables. Since some patients had to be excluded and may be more able to improve beta and from the matched cohorts because of missing alpha cell function . values in secondary outcome variables, results These ﬁndings have clinical implications for for FPG, body weight, and blood pressure have individualization of therapy based on patients’ to be considered with more caution than results history, indicating that DPP4i can be effective for HbA1c. Finally, since only a fraction of the also in a more advanced disease stage. initial patient cohort could be matched, results Although DPP4i are associated with a mark- apply only to patients with the baseline clinical edly lower risk of hypoglycemia than SUs, characteristics obtained after PSM. information on hypoglycemic events are not On the other hand, the study has remarkable yet available in the DARWIN-T2D database and strengths. These include the large sample size, it is therefore impossible to weigh the beneﬁts the extensive patient characterization, the of glucose control against the risk of hypo- multicenter nature with nationwide distribu- glycemia. Future real-world studies combining tion, the rigorous consideration of biases, and clinical and administrative data on hospital the automatic data extraction from the same discharge codes will be useful to address this electronic chart, which guarantees repro- issue. ducibility, uniform data coding, and low In addition to the glucose-lowering potency reporting bias. and the risk of hypoglycemia, there is great focus on cardiovascular effects of drugs for the CONCLUSION treatment of T2D. The ongoing CAROLINA trial, comparing linagliptin versus glimepiride, Addition of a DPP4i to an ongoing glucose- will shed light on cardiovascular outcomes with lowering regimen in Italian diabetes specialist DPP4i versus SUs . outpatient clinical practice improved glucose This study has limitations inherent to its control more than addition of gliclazide. retrospective and non-randomized design. Although gliclazide was being used at submax- Therefore, the level of evidence arising from imal doses and confounding cannot be deﬁ- these data cannot be equated to that of RCTs. nitely ruled out, these data confute the general The risk of confounding by indication and belief that initiation of SUs is highly effective in reverse causality always limit interpretation of reducing HbA1c and provide a rationale for the comparisons between therapeutic strategies pragmatic trials comparing DPP4i and SU in a in observational studies. To address this issue, routine clinical setting. we used PSM to obtain matched cohorts of patients and simulate a quasi-experimental design. With this tool, it is possible to emulate the conditions of an RCT with respect to the ACKNOWLEDGEMENTS observed baseline characteristics. By PSM, we have been able to obtain well-balanced groups, We wish to thank the technical support of except for a residual difference in the rate of Alessia Russo, Italian Diabetes Society. concomitant basal insulin use, at a magnitude that may not be clinically relevant. Nonethe- Funding. The DARWIN-T2D study was fun- less, to account for this residual confounding, ded by the Italian Diabetes Society, with exter- the outcome analysis was adjusted for insulin nal support from AstraZeneca. No funding or use or presented separately for insulin users and 1488 Diabetes Ther (2018) 9:1477–1490 sponsorship was received for the publication of the institutional and/or national research com- this article and the external funding source had mittee and with the 1964 Declaration of Hel- no role in study design and conduction, nor in sinki and its later amendments or comparable the writing of the manuscript and decision to ethical standards. Since the study was per- publish. formed retrospectively on an anonymized database, no patient consent was required. Authorship. All named authors meet the International Committee of Medical Journal Data Availability. The data sets generated Editors (ICMJE) criteria for authorship for this during and/or analyzed during the current article, take responsibility for the integrity of study are available from the corresponding the work as a whole, and have given their author on reasonable request. approval for this version to be published. Composition of the DARWIN-T2D Data- Authorship Contributions. Design, data base. Agostino Consoli and Gloria Formoso collection and analysis: Gian Paolo Fadini, (Dipartimento di Medicina e Scienze dell’In- Giancarlo Zatti, Daniele Bottigliengo, Ileana vecchiamento - Universita ` Degli studi G. D’An- Baldi, Federica D’Angelo, Franco Cavalot, nunzio di Chieti-Pescara); Giovanni Grossi Antonio Carlo Bossi and Angelo Avogaro. (Ospedale San Francesco di Paola - Azienda Manuscript writing: Gian Paolo Fadini and Sanitaria Provinciale di Cosenza); Achiropita Angelo Avogaro. Manuscript revising for intel- Pucci (Azienda Sanitaria Provinciale di Cosen- lectual content: Gian Paolo Fadini, Daniele za); Giorgio Sesti and Francesco Andreozzi Bottigliengo, Ileana Baldi, Federica D’Angelo, (Azienda Ospedaliero Universitaria di Catan- Franco Cavalot, Antonio Carlo Bossi. zaro); Giuseppe Capobianco (Azienda Sanitaria Locale Napoli 2 Nord); Adriano Gatti (Ospedale Disclosures. Gian Paolo Fadini received San Gennaro dei Poveri - Azienda Sanitaria grant support, lecture or advisory board fees Locale Napoli 1 Centro); Riccardo Bonadonna, from AstraZeneca, Boehringer-Ingelheim, Eli Ivana Zavaroni and Alessandra Dei Cas (Azienda Lilly, NovoNordisk, Sanoﬁ, Genzyme, Abbott, Ospedaliero Universitaria di Parma); Giuseppe Novartis, Merck Sharp & Dohme. Federica Felace (Ospedale di Spilimbergo - Azienda per D’Angelo received grants from AstraZeneca, l’Assistenza Sanitaria n.5 Friuli Occidentale); Novonordisk and Boehringer-Ingelheim. Anto- Patrizia Li Volsi (Ospedale di Pordenone - nio Carlo Bossi received grants or personal fees Azienda per l’Assistenza Sanitaria n.5 Friuli from Lilly, NovoNordisk, Johnson & Johnson, Occidentale); Raffaella Buzzetti and Gaetano Boehringer-Ingelheim, Artsana, Takeda, Bayer, Leto (Ospedale Santa Maria Goretti - Azienda Sanoﬁ, AstraZeneca. Franco Cavalot received Sanitaria Locale di Latina); Gian Pio Sorice grant support or lecture fees from AstraZeneca, (Fondazione Policlinico Universitario A. Sanoﬁ, NovoNordisk, Boeringher-Ingelheim Gemelli, Roma); Paola D’Angelo (Ospedale and Takeda. Angelo Avogaro received research Sandro Pertini - Azienda Sanitaria Locale Roma grants, lecture or advisory board fees from 2); Susanna Morano (Azienda Ospedaliera Merck Sharp & Dome, AstraZeneca, Novartis, Universitaria Policlinico Umberto I, Roma); Boeringher-Ingelheim, Sanoﬁ, Mediolanum, Antonio Carlo Bossi (Ospedale di Treviglio - Janssen, NovoNordisk, Lilly, Servier, and Azienda Socio Sanitaria Territoriale Bergamo Takeda. Giancarlo Zatti, Daniele Bottigliengo Ovest); Edoardo Duratorre (Ospedale Luini and Ileana Baldi have nothing to disclose. Confalonieri di Luino - Azienda Socio Sanitaria Territoriale Sette Laghi). Ivano Franzetti (Ospe- Compliance with Ethics Guidelines. The dale Sant’Antonio Abate di Gallarate - Azienda study was approved by the ethical committee of Socio Sanitaria Territoriale Valle Olona); Paola each participating center. All procedures per- Silvia Morpurgo (Ospedale Fatebenefratelli - formed in studies involving human participants Azienda Socio Sanitaria Territoriale Fatebene- were in accordance with the ethical standards of fratelli Sacco); Emanuela Orsi (Fondazione Diabetes Ther (2018) 9:1477–1490 1489 IRCCS Ca’ Granda - Ospedale Maggiore Poli- San Bonifacio - Azienda ULSS n.9 Scaligera); clinico di Milano); Fabrizio Querci (Ospedale Natalino Simioni and Vera Frison (Ospedale di Pesenti Fenaroli di Alzano Lombardo - Azienda Cittadella - Azienda ULSS n.6 Euganea); Car- Socio Sanitaria Territoriale Bergamo Est); Mas- mela Vinci (Azienda ULSS n.4 Veneto simo Boemi and Federica D’Angelo (Presidio Orientale). Ospedaliero di Ricerca INRCA-IRCCS di Open Access. This article is distributed Ancona); Massimiliano Petrelli (Azienda Ospe- under the terms of the Creative Commons daliero Universitaria Ospedali Riuniti di Attribution-NonCommercial 4.0 International Ancona); Gianluca Aimaretti and Ioannis Kar- License (http://creativecommons.org/licenses/ amouzis (Azienda Ospedaliero Universitaria by-nc/4.0/), which permits any noncommercial Maggiore della Carita di Novara); Franco Cava- use, distribution, and reproduction in any lot (Azienda Ospedaliero Universitaria San Luigi medium, provided you give appropriate credit to Gonzaga, Orbassano); Giuseppe Saglietti the original author(s) and the source, provide a (Ospedale Madonna del Popolo di Omegna - link to the Creative Commons license, and Azienda Sanitaria Locale Verbano Cusio Ossola); indicate if changes were made. Giuliana Cazzetta (Casa della Salute, Ugento - Distretto Socio Sanitario Gagliano del Capo - Azienda Sanitaria Locale di Lecce); Silvestre Cervone (Presidio ospedaliero San Marco in REFERENCES Lamis - Distretto Socio Sanitario San Marco in Lamis - Azienda Sanitaria Locale di Foggia); 1. Inzucchi SE, Bergenstal RM, Buse JB, et al. Man- Eleonora Devangelio (Distretto Socio Sanitario agement of hyperglycemia in type 2 diabetes, 2015: di Massafra - Azienda Sanitaria Locale di Tar- a patient-centered approach: update to a position anto); Olga Lamacchia (Azienda Ospedaliero statement of the American Diabetes Association Universitaria Ospedali Riuniti di Foggia); Salva- and the European Association for the Study of Diabetes. Diabetes Care. 2015;38:140–9. tore Arena (Ospedale Umberto I - Azienda San- itaria Provinciale di Siracusa); Antonino Di 2. Frieden TR. Evidence for health decision making— Benedetto (Azienda Ospedaliera Universitaria beyond randomized, controlled trials. N Engl J Med. Policlinico G. Martino di Messina); Lucia Frit- 2017;377:465–75. titta (Azienda Ospedaliera di Rilievo Nazionale e 3. Mishriky BM, Cummings DM, Tanenberg RJ. The di Alta Specializzazione Garibaldi di Catania); efﬁcacy and safety of DPP4 inhibitors compared to Carla Giordano (Azienda Universitaria Policlin- sulfonylureas as add-on therapy to metformin in ico Paolo Giaccone di Palermo); Salvatore Piro patients with type 2 diabetes: a systematic review and meta-analysis. Diabetes Res Clin Pract. (Azienda Ospedaliera di Rilievo Nazionale e di 2015;109:378–88. Alta Specializzazione Garibaldi di Catania); Manfredi Rizzo, Roberta Chianetta and Carlo 4. Foroutan N, Muratov S, Levine M. Safety and efﬁ- Mannina (Azienda Universitaria Policlinico cacy of dipeptidyl peptidase-4 inhibitors vs sul- fonylurea in metformin-based combination therapy Paolo Giaccone di Palermo); Roberto Anichini for type 2 diabetes mellitus: systematic review and (Ospedale San Jacopo di Pistoia - Azienda USL meta-analysis. Clin Invest Med. 2016;39:E48–62. Toscana Centro); Giuseppe Penno (Azienda Ospedaliero Universitaria Pisana); Anna Solini 5. Green JB, Bethel MA, Armstrong PW, et al. Effect of sitagliptin on cardiovascular outcomes in type 2 (Azienda Ospedaliera Universitaria Pisana); diabetes. N Engl J Med. 2015;373:232–42. Bruno Fattor (Comprensorio Sanitario di Bol- zano - Azienda Sanitaria della Provincia Auton- 6. White WB, Cannon CP, Heller SR, et al. Alogliptin oma di Bolzano); Enzo Bonora and Massimo after acute coronary syndrome in patients with type 2 diabetes. N Engl J Med. 2013;369:1327–35. Cigolini (Azienda Ospedaliero Universitaria Integrata di Verona); Annunziata Lapolla and 7. Scirica BM, Bhatt DL, Braunwald E, et al. Saxagliptin Nino Cristiano Chilelli (Complesso Socio Sani- and cardiovascular outcomes in patients with type tario Ai Colli - Azienda ULSS n.6 Euganea); 2 diabetes mellitus. N Engl J Med. 2013;369:1317–26. Maurizio Poli (Ospedale Girolamo Fracastoro di 1490 Diabetes Ther (2018) 9:1477–1490 8. Fadini GP, Avogaro A, Degli Esposti L, et al. Risk of 14. Fadini GP, Zatti G, Consoli A, et al. Rationale and hospitalization for heart failure in patients with design of the DARWIN-T2D (DApagliﬂozin Real type 2 diabetes newly treated with DPP-4 inhibitors World evIdeNce in Type 2 Diabetes): a multicenter or other oral glucose-lowering medications: a ret- retrospective nationwide Italian study and crowd- rospective registry study on 127,555 patients from sourcing opportunity. Nutr Metab Cardiovasc Dis. the Nationwide OsMed Health-DB Database. Eur 2017;27:1089–97. Heart J. 2015;36:2454–62. 15. Friedewald WT, Levy RI, Fredrickson DS. Estimation 9. Tzoulaki I, Molokhia M, Curcin V, et al. Risk of of the concentration of low-density lipoprotein cardiovascular disease and all cause mortality cholesterol in plasma, without use of the prepara- among patients with type 2 diabetes prescribed oral tive ultracentrifuge. Clin Chem. 1972;18:499–502. antidiabetes drugs: retrospective cohort study using UK general practice research database. BMJ. 16. Levey AS, Stevens LA, Schmid CH, et al. A new 2009;339:b4731. equation to estimate glomerular ﬁltration rate. Ann Intern Med. 2009;150:604–12. 10. Chan SP, Colagiuri S. Systematic review and meta- analysis of the efﬁcacy and hypoglycemic safety of 17. Rosenbaum PR, Rubin DB. The central role of the gliclazide versus other insulinotropic agents. Dia- propensity score in observational studies for causal betes Res Clin Pract. 2015;110:75–81. effects. Biometrika. 1983;70:41–55. 11. Schramm TK, Gislason GH, Vaag A, et al. Mortality 18. Rubin DB. Multiple imputation for nonresponse in and cardiovascular risk associated with different survey. Wiley Online Library. 1987. insulin secretagogues compared with metformin in type 2 diabetes, with or without a previous 19. Rubin RB. Multiple imputation after 18? years. myocardial infarction: a nationwide study. Eur J Am Stat Assoc. 1996;91:473–89. Heart J. 2011;32:1900–8. 20. Derosa G, Cicero AF, Franzetti IG, et al. A ran- 12. Fadini GP, Zatti G, Baldi I, et al. Use and effective- domized, double-blind, comparative therapy eval- ness of dapagliﬂozin in routine clinical practice. An uating sitagliptin versus glibenclamide in type 2 Italian multicenter retrospective study. Diabetes diabetes patients already treated with pioglitazone Obes Metab. 2018. https://doi.org/10.1111/dom. and metformin: a 3-year study. Diabetes Technol 13280. Ther. 2013;15:214–22. 13. Vaccaro O, Masulli M, Nicolucci A, et al. Effects on 21. Marx N, Rosenstock J, Kahn SE, et al. Design and the incidence of cardiovascular events of the addi- baseline characteristics of the CARdiovascular tion of pioglitazone versus sulfonylureas in patients Outcome Trial of LINAgliptin Versus Glimepiride in with type 2 diabetes inadequately controlled with Type 2 Diabetes (CAROLINA). Diab Vasc Dis Res. metformin (TOSCA.IT): a randomised, multicentre 2015;12:164–74. trial. Lancet Diabetes Endocrinol. 2017;5:887–97.
– Springer Journals
Published: Jun 1, 2018