gnificant heterogeneity was considered if P <0.1. I2 statistic was also examined, where I2 of 25%, 50%, and 75% indicated low, moderate, and high degrees of heterogeneity, respectively. Publication bias was assessed by using funnel plots and Egger's test. The software RevMan and Stata supported the analysis. Subgroup analysis was performed on the basis of the types of trial drug. Results Search results A total of 2,377 studies were initially identified; among these studies, 2,320 were excluded after titles and abstracts were screened. The full texts of the 57 remaining studies were analyzed. Among these 57 studies, 30 were excluded because of the following reasons: first, 11 studies provided insufficient data on related outputs; second, 5 studies did not include an appropriate control group; third, 3 studies used an ineligible study design; fourth, participants in 7 studies were not patients with type 2 diabetes. Finally, the endpoints were not relevant in 3 studies and 1 study reported replicated data. Study characteristics A total of 27 RCTs, including 5 open-labeled randomized trials, were eligible for the present meta-analysis. 11 of the observational studies were used rosiglitazone as the primary source of TZDs; in the other studies, pioglitazone was used. Pioglitazone and rosiglitazone doses varied from 15 mg/d to 45 mg/d and from 4 mg/d to 8 mg/d, respectively. Among the 27 studies, 7 investigated the effects of TZDs on type 2 diabetic patients with coronary 3 / 15 Inflammatory Markers in Type 2 Diabetes Fig 1. Process of study selection. doi:10.1371/journal.pone.0123703.g001 artery disease. 12 studies examined type 2 diabetic subjects only, and the remaining studies were conducted on individuals with a combination of type 2 diabetes and one of the following conditions: dislipidemia; obesity; asymptomatic PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19769788 carotid stenosis; chronic kidney disease; atherosclerosis and kidney transplant. 4 / 15 Inflammatory Markers in Type 2 Diabetes study NO. of patients exposed group Bertrand 2010 Finn 2009 Schondorf 2011 Harteman 2009 Kiyici 2009 98 32 25 14 control group 95 33 21 13 mean age exposed group 64.27.3 65.79.6 59.48.0 6210 control group 65.1 6.9 59.7 9.3 57.5 10.1 5810 T2DM with CAD T2DM with CAD T2DM with dislipidimia T2DM with obesity T2DM population dosage mg/d 8 4 15 30 12m 8m 24w 24w Ra, DB PC Ra, PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19768500 DB, PC Ra,DB, PC Ra, PC R R P P R not TZDs use metformin Glimepiride insule metformin 5 4 3 5 duration design Medication studied J score 19 16 50.76.4 52.4 8.3 65.7 7.5 5310.4 52.6 7.2 683 4 52w Ra, PCopenlabel Ra,DB, PC Ra,DB, PC Ra,PC Ra,DB, PC 3 Forst 2008 Davidson 2007 Fidan 2011 Marfella 2006 44 117 20 23 48 116 20 23 64.57.3 5211.9 54.19.0 692 T2DM with CAD T2DM T2DM T2DM with asymptomatic carotid stenosis T2DM with 
chronic kidney disease T2DM 45 8 4/8 8 4w 24w 12w 4m P R R R R not TZDs use not TZDs use metformin not TZDs use not TZDs use 3 4 3 3 Chan 2011 35 35 6210 6210 4 8w Ra,DB, PC Ra,PC openlabel Ra,PC Ra,SB, PC Ra,PC openlabel Ra,PC Ra,DB, PC Ra,PC Ra,DB, PC Ra,DB, PC Ra,PC 4 Pfuttzner 2005 Mizoguchi 2011 Hong 2010 SAR 405 chemical information Oqasawara 2009 Yu 2007 Hanefeld 2011 Yener 2009 Kelly 2007 Nissen 2008 Langenfeld 2005 89 84 62.28.4 63.0 7.4 68.0 9.1 62.4 8.3 66.8 8.1 63.5 9.8 64.2 7.3 53.2 8.0 63.1 7.8 59.7 9.1 637 45 262w P glimepiride 3 31 47 22 21 47 24 68.27.3 63.57.4 68.67.9 T2DM with atherosclerosis T2DM with CAD T2DM with CAD 30 30 15 4m 8m 6m P P P glimepiride not TZDs use not TZDs use 3 3 3 25 40 20 20 270 89 21 42 20 16 273 84 63.6