Table 1. Study characteristics.

Pavo et al.,2003 [9], Schernthaner et al.,2004 [10], Fidan et al.,2011 [11], Chu et al.,2002 [12], Hällsten et al.,2002 [13], Kautzky-Willer et al.,2005 [14], Tiikkainen et al.,2004 [15], Ceriello et al.,2005 [16], Roden et al.,2005 [17], Basu et al.,2008 [18]

Author ID	Study Design	Participant characteristics	Metformin	TZDs	Follow-up (weeks)	Main outcomes
Pavo et al.,2003 [9]	RCT	205 patients (103 females and 102 males, aged at least 40 years)	100 patients in the metformin group received a mean dosage of 2292 mg/d	105 patients in the pioglitazone group received a mean dosage of 41.5 mg/d	32	The patients in the pioglitazone group had a more pronounced HOMA-S than those in the metformin group (mean treatment difference, 16.37; SD, 6.77; P < 0.05). There was a significant difference in the reduction of FPG between the two groups (-3.0 mmol/l vs. -2.8 mmol/l, p = 0.620 for pioglitazone and metformin, respectively).
Schernthaner et al.,2004 [10]	RCT	1194 patients (659 males and 535 females)	65, 146, and 339 patients were subjected to metformin dosages of 850mg, 1700mg, and 2550mg, respectively.	78 and 475 patients were subjected to 30mg and 45mg of pioglitazone, respectively.	52	The fasting insulin was significantly reduced in the pioglitazone group than in the metformin group (-2.4 µU/ml vs. 0 µU/ml, p<0.0001, respectively). The Mean change in the FPG was larger in patients receiving pioglitazone than in patients receiving metformin (-45.0±16.2 mg/dL vs. -39.6±16.2 mg/dL, respectively). 42 patients receiving pioglitazone and 39 receiving metformin were withdrawn from the study due to adverse events.
Fidan et al.,2011 [11]		40 patients (16 females and 24 males aged above 40 years)	Metformin was initiated at 850 mg dosage and adjusted to 3 x 850 mg at 10-day intervals	Rosiglitazone was initiated at 4mg and adjusted to 8 mg.	12	At the end of 3 months, the decrease in HOMA-IR was not significant for patients receiving metformin (from 4.5±3.0 to 3.4±3.5), while a significant difference was observed in the rosiglitazone group (from 4.8 ± 3.4 to 2.8 ± 2.5, p<0.05). An insignificant difference in the FPG was observed in the metformin group (from 144 ± 24mg/dL to 130 ± 36mg/dL), while a significant change was recorded in the rosiglitazone group (from 149 ± 30mg/dL to 116 ± 20mg/dL p<0.05).
Chu et al.,2002 [12]	RCT	22 patients (20 males and 2 females; mean age 56±2 years).	12 patients were initially subjected to receive 850mg daily	10 patients were initially subjected to 200mg troglitazone daily.	16	An insignificant change in the FPG was recorded between the metformin and troglitazone groups (-32mg/dL vs. -36mg/dL, p=0.90). Both groups’ fasting insulin levels decreased; however, the difference was insignificant (-6mU/l vs. -18mU/l, p=0.30, for metformin and troglitazone, respectively).
Hällsten et al.,2002 [13]	RCT	41 patients (28 males and 13 females)	13 patients were initially subjected to 500mg of metformin for 2 weeks and thereafter 1g daily	14 patients were initially subjected to 2mg of rosiglitazone for 2 weeks; thereafter, 4mg daily	26	A significant change in FPG was recorded in the metformin group (from 8.0±0.5mmol/l to 6.8±0.3mmol/l, p < 0.001), while an insignificant decrease in FPG was recorded in the rosiglitazone group (from 7.2±0.3 to 6.8±0.3mmol/l). The decrease in fasting serum insulin was insignificant in both metformin (from 11.7± 2.1mU/l to 8.8±1.1mU/l) and rosiglitazone group (from 8.6±1.5 to 6.6±0.4).
Kautzky-Willer et al.,2005 [14]	RCT	20 patients (14 males and 6 females)	9 patients were initially subjected to an 850mg metformin dose once every day for 1 week; thereafter, the dose was given twice daily.	11 patients were initially subjected to a 400mg daily dosage of troglitazone for 1 week; thereafter, the dose was adjusted to 600mg daily.	16	A significant improvement in fasting insulin resistance (HOMA) was recorded in the troglitazone group (from 5.3±0.9 to 3.4±0.7), while the change was insignificant for the metformin group (from 6.3±0.9 to 4.1±0.8, respectively).
Tiikkainen et al.,2004 [15]		20 patients (13 females and 7 males)	11 patients in the metformin group received 1g dosage for 16 weeks	9 patients were subjected to 4mg of rosiglitazone for 16 weeks	16	Fasting serum insulin decreased significantly in both groups; however, the statistical analysis showed that the difference was not significant (4±1 and 4±2mU/l for rosiglitazone and metformin groups, respectively) A significant decrease in FPG was recorded in rosiglitazone (from 8.8±0.8 to 7.3±0.4mmol/l) and metformin group (from 8.2±0.7 to 6.7±0.2mmol/l)
Ceriello et al.,2005 [16]	RCT	940 patients (382 females and 558 males)	195 patients were subjected to pioglitazone	187 patients received metformin	52	There was a statistically insignificant difference in the change of FPG between pioglitazone and metformin (-2.52±0.145 vs.  -2.46±0.148mmol/l, respectively). Pioglitazone had a significantly higher reduction in fasting insulin than metformin (-12.77±2.40 vs. -4.59±2.44mmol/l).
Roden et al.,2005 [17]	RCT	1788 patients (1005 males and 783 females aged 35 – 75 years)	597 patients received metformin	597 patients received pioglitazone	52	A higher significant increase in insulin sensitivity was recorded among patients in the pioglitazone group than in the metformin group (p<0.001). A significantly higher reduction in fasting serum insulin was recorded among patients in the pioglitazone group (from −16.12±3.15 to −27.38 ± 2.17pmol/l).
Basu et al.,2008 [18]	RCT	31 patients (15 females and 16 males)	16 patients were subjected to 1000mg of metformin twice every day	15 patients were subjected to 45mg of pioglitazone daily	16	No significant change in fasting glucose was recorded in either patients receiving pioglitazone (from 157 ± 10 to 140 ± 13mg/dL) or metformin (from 148 ± 12 to 146 ± 10mg/dL).