Table 2.

Main characteristics of studies included in the meta-analysis

Reference, source population, study period	Study design, population size, age	Diabetes type 2 population definition	Study endpoints (number of cases)	Case validation	Exposure assessment	Exposure recency	Exposure group(s) vs. reference group (n) A: Comparison(s) contributing to meta-analysis B: Other comparison(s)
Studies included in both meta-analysis endpoints, AMI and stroke
Bilik, 2010 [26] TRIAD study group USA 2000–2003	Cohort N = 2,382 Age ≥ 18 years (patients taking only insulin and aged younger than 30 years were excluded)	First prescription for glitazones. Patients filling prescriptions for more than one TZD were excluded	Non-fatal AMI (ICD-9: 410) (N = 39) Non-fatal stroke (ICD-9: 431, 433, 434) (N = 32)	None	Prevalent and new users Dispensed prescriptions	Current, continuous use until 90 days after the supply date of their most recently filled prescription duration	A: Rosiglitazone (n = 773) vs. pioglitazone (n = 711)
Graham, 2010 [30] Medicare, USA 2006–2009	Cohort 227,571 Age ≥ 65 years	First prescriptions for glitazones	Hospitalisation for fatal and non-fatal AMI/ACS/SCHD (ICD-9: 410) (N = 1,746) Hospitalisation for fatal and non-fatal acute stroke, ischaemic or haemorrhagic (ICD-9: 430, 431, 433.x1, 434.x1, and 436) (N = 1,052)	External PPV for AMI: 89 % – 97 % PPV for stroke: 92 % – 100 %	New users Dispensed prescriptions	Current continuous use, including a gap of no more than 7 days	A: Rosiglitazone (n = 67,593) vs. pioglitazone (n = 159,978)
Winkelmayer, 2008 [41] Medicare, New Jersey, USA 1999–2005	Cohort 28,361 Age > 65 years	First prescription for a glitazone, regardless of previous treatment with other diabetic drug(s)	Ever and first ever Fatal and non-fatal AMI (ICD-9: not reported ) (n = 737) Hospitalisation for a non-fatal stroke: ischaemic or haemorrhagic (ICD-9: 433,434,436) (N = 1,869)	External PPV for AMI: 94 % PPV for stroke: 96 %	New users Dispensed prescriptions	Current, continuous use until 60 days after the end of supply date of their most recently filled prescription duration or until switching to other TZD	A: Rosiglitazone (n = 14,101) vs. pioglitazone (n = 14,260)
Studies included only in the meta-analysis of AMI
Horsdal, 2011 [31] Danish National Registries 1996–2004	Nested case–control N = 101,313 >30 years	Subjects were classified as having T1DM and excluded if they were aged younger than 30 years at the time of their first related prescription or diagnosis and had never received a prescription for an oral glucose–lowering drug. Subjects with T2DM were those with codes for diabetes mellitus who had not received pharmacotherapy, had received prescriptions for oral glucose–lowering drugs, or were aged older than 30 years when they had their first diagnostic code or prescription.	Hospitalisation for AMI (codes not reported) (N = 10,616)	External	Prevalent and new users Dispensed prescriptions	At least one prescription of study drug within 90 days before hospitalisation	A: Sulfonylurea monotherapy (n = 26,778) vs. metformin monotherapy (n = 5,927) B: Sulfonylurea monotherapy (n = 26,778) vs. any combination (n = 12,425); metformin monotherapy (n = 5,927) vs. individual sulfonylurea monotherapy (n = 26,778)
Loebstein, 2011 [35] Maccabi Healthcare Services, Israel 2000–2007	Cohort N = 15,436 Age, mean (SD): 59.1 (11.4) years	Subjects in the Maccabi diabetes registry with prescriptions for rosiglitazone or metformin for at least 6 months	Hospitalisation for AMI (ICD-9 and Y codes 410XX, Y139XX and Y225XX) (N = 645)	None	Prevalent and new users Dispensed prescriptions	Current, continuous use within study period with gaps not longer than 3 months	A: Rosiglitazone monotherapy (n = 745) or in combination with metformin (n = 2,753) vs. metformin monotherapy (n = 11,938) (Formulary restriction allowed to use rosiglitazone only if inadequate control from SU, metformin, or both)
Brownstein, 2010 [27] Partners Healthcare System: Research Patient Data Registry, USA 2000–2006	Cohort N = 26,375 Age ≥ 18 years	ICD-9: 250.XX or hemoglobin A1C of at least 6.0 % and at least one record of prescription of an oral diabetes medication as an outpatient or dispensing as an inpatient	Hospitalisation for AMI (ICD-9: 410) (N = 1,343)	External PPV: 92 % –94 %	Prevalent and new users Prescriptions issued and dispensed	Current, continuous use within study period with gaps not longer than 6 months	A: Rosiglitazone monotherapy (n = 1,879) vs. pioglitazone monotherapy (n = 806) or metformin monotherapy (n = 12,490) or sulfonylurea monotherapy (n = 11,200)
Wertz, 2010 [40] HealthCore Integrated Research Database, USA 2001–2005	Cohort N = 36,628 Age ≥ 18 years	First prescription for glitazones	Hospitalisation for AMI (ED visits included) (ICD-9 410.xx) (n = 217)	None	New users Dispensed prescriptions	Current use if refill occurred < 1.5 times the days’ supply of the preceding claim for TZD	A: Rosiglitazone (n = 14,469) vs. pioglitazone (n = 14,469)
Dormuth, 2009 [28] British Columbia Health databases, Canada 1997–2007	Nested case–control N = 11,147 Age, mean (SD): 70 (12) years	Subjects with a pharmacy dispensing for metformin, without a dispensing of metformin, other antidiabetic medication, or insulin in the previous 365 days	Hospitalisations for fatal and non-fatal AMI (ICD-9: 410) (N = 2,244)	None	New users Dispensed prescriptions	Current use within 90 days of the index date	A: Rosiglitazone (n = 462) vs. pioglitazone (n = 235) Rosiglitazone (n = 462) vs. metformin (n = 10,685) Rosiglitazone (n = 462) vs. sulfonylurea (n = 1,612) Pioglitazone (n = 235) vs. metformin (n = 10,912) Sulfonylurea (n = 1,612) vs. metformin (n = 9,535)
Hsiao, 2009 [32] Taiwan Health Insurance Database 2001–2005	Cohort N = 473,483 Age, not reported	Subjects with their first ambulatory visit with ICD-9-CM code 250.xx who were prescribed oral blood glucose–lowering agents at least three times. Subjects were excluded if they had T1DM (ICD-9-CM codes 250.x1) or if they had been prescribed only insulin during the study period.	Fatal and non-fatal hospitalisation for AMI (ICD-9: 410.xx and 411.xx) (N = 15,917)	None	New users Dispensed prescriptions	Current, continuous use during study period	A: Pioglitazone monotherapy (n = 495) or rosiglitazone monotherapy (n = 2,093) vs. metformin-based therapy (n = 46,444) and vs. SU-based therapy (n = 97,651) B: Pioglitazone + SU + metformin (n = 9,510) vs. Rosiglitazone + SU + metformin (n = 39,962) Pioglitazone + metformin (n = 774) vs. rosiglitazone + metformin (n = 2,408) Pioglitazone + SU (n = 1,231) vs. rosiglitazone + SU (n = 5,141)
Juurlink, 2009 [33] Ontario diabetes database, Canada 2002–2008	Cohort N = 39,736 Age ≥ 66 years	First prescription for a glitazone	Hospitalisation for AMI (ICD-10: I20, I21, I22) (N = 698)	External PPV ≈ 90 %	New users prescription Dispensed prescriptions	Current use if refill occurred < 1.5 times the days’ supply of the preceding TZD claim	A: Rosiglitazone (n = 22,785) vs. pioglitazone (n = 16,951)
Tzoulaki, 2009 [38] GPRD, United Kingdom 1990–2005	Cohort N = 91,521 Age 35–90 years	One episode of care associated with a clinical or referral event for diabetes and prescriptions for oral blood glucose–lowering treatment	First ever diagnosis of AMI according to Read codes (N = 3,588)	External Confirmed 90 % of AMI diagnoses	Prevalent and new users Prescriptions issued	Current, continuous intervals of use within the study period	A: First-generation SU monotherapy (n = 6,053) or second-generation SU monotherapy (n = 58,095) or rosiglitazone monotherapy (n = 8,442) and combination therapy (n = 9,640) or pioglitazone including monotherapy and combination therapy (n = 3,816) vs. metformin (n = 68,181) B: Glibenclamide or gliclazide or glimepiride or glipizide or gliquidone vs. metformin (n = 68,181)
Ziyadeh, 2009 [42] i3, USA 2000–2007	Cohort N = 95,002 Age ≥ 18 years	Initiators of glitazones	Hospitalisations for fatal and non-fatal AMI (ICD-9: 410.xx) (N = 460)	External	New users Dispensed prescriptions	Current, use at index date	A: Rosiglitazone monotherapy (n = 47,501) vs. pioglitazone monotherapy (n = 47,501)
Koro, 2008 [34] Integrated HealthCore Information Services, USA 1999–2006	Nested case–control N = 891,901 Age ≥ 30 years	Subjects with a diagnosis of type 2 diabetes and at least one prescription claim for an antidiabetic agent during their follow-up time available in the database	First-ever hospitalisation for AMI (ICD-9: 410.xx) (N = 9,870)	None	Prevalent and new users Dispensed prescriptions	Current use, a prescription in the last 3 months prior to index date	A: Rosiglitazone (n = 3,839) vs. pioglitazone (n = 3,343)
Walker, 2008 [39] Pharmetrics integrated outcomes database USA 2000–2007	Cohort ≈543,000 Age ≥ 18 years	Subjects were users of rosiglitazone, pioglitazone, metformin, or a sulfonylurea	Hospitalisation for fatal and non-fatal AMI (no codes reported) (N = 502)	None	New users Dispensed prescriptions	Current, use at index date	A: Rosiglitazone monotherapy (n = 12,440) vs. pioglitazone monotherapy (n = 16,302); rosiglitazone monotherapy (n = 12,440) or pioglitazone monotherapy (16,302) vs. metformin (n = 131,075); rosiglitazone monotherapy (n = 12,440) vs. sulfonylurea monotherapy (n = 48,376)
Gerrits, 2007 [29] Ingenix Research Database, USA 2003–2006	Cohort N = 29,911 Age ≥ 45 years	Subjects with ICD-9 code of 250.xx and a dispensing of pioglitazone or rosiglitazone	Hospitalisation for AMI (ICD-9: 410.xx) (N = 375)	External; PPV ≥ 95 %	New users Dispensed prescriptions	Exposure to pioglitazone and rosiglitazone was treated as a unidirectional time-varying covariate; that is, once a patient met the exposure definition, the patient was considered exposed from that point forward, even if the index drug was discontinued	A: Rosiglitazone (n = 15,104) vs. pioglitazone (n = 14,807)
McAfee, 2007 [36] Ingenix Research Database, USA 2000–2004	Cohort N = 33,363 Age ≥ 18 years	Initiators of rosiglitazone, metformin, or a sulfonylurea	Hospitalisation for AMI (ICD-9: 410.xx) (N = 226)	External	New users Dispensed prescriptions	Current use during study period. Dispensing of a different study drug or insulin for the monotherapy group (at which time the subject became eligible for a different study cohort); cessation of study drug use alone was not sufficient to end follow-up	A: Rosiglitazone monotherapy (n = 8,977) vs. metformin monotherapy (n = 8,977) or sulfonylureas monotherapy (n = 8,977) B: Rosiglitazone + metformin (n = 1,362) or rosiglitazone + sulfonylurea (n = 1,362) vs. metformin + sulfonylurea (n = 1,362) Rosiglitazone (n = 12,874) vs. non-rosiglitzazone drug (n = 20,489)
Sauer, 2006 [37] Philadelphia Metropolitan area, USA 1998–2002	Case–control (field study) Controls were community controls selected using random digit dialing N = 764 Age 40–75 years	Subjects with T2DM treated with antidiabetic drugs or diet only	First-ever AMI (identified using medical records) (N = 113)	AMI validated by Minnesota Heart Survey criteria	Prevalent and new users Interviews	Current, use in the 7 days before index date	A: Sulfonylurea monotherapy (n = 158) vs. metformin monotherapy (n = 125) B: Sulfonylureas (n = 158) or metformin (125) vs. thiazolidinedione (n = 26) Sulfonylureas (n = 158) vs. thiazolidinediones + sulfonylureas (n = 27) Sulfonylureas (n = 158) vs. metformin + sulfonylureas (n = 102) Metformin (n = 125) vs. thiazolidinediones + metformin (n = 21)

ACS acute coronary syndrome, AMI acute myocardial infarction, ED emergency department, GPRD General Practice Research Database (now the Clinical Practice Research Datalink [CPRD]), hemoglobin A1C glycated hemoglobin, ICD-9 international classification of diseases, 9th revision, ICD-9-CM, International Classification of Diseases, 9th Revision, Clinical Modification; ICD-10, International Statistical Classification of Diseases and Related Health Problems, 10th Revision; PPV, positive predictive value; SCHD, serious coronary heart disease; SD, standard deviation; SU, sulphonylurea(s); T1DM, type 1 diabetes mellitus; T2DM type 2 diabetes mellitus, TZD thiazolidinedione(s), USA United States of America

Note: When it is not indicated that the endpoint is the first ever identified, the study included patients with and without prior history of the study endpoint