Table 4: Risk-of-bias assessment of the trials included in the meta-analysis* .

Trial	Sequence generation (selection bias)	Allocation concealment (selection bias)	Blinding of participants and personnel (performance bias)	Blinding of outcome assessors (detection bias)	Completeness of outcome data (attrition bias)	Selective reporting (reporting bias)	Academic bias	Sponsor bias
ADOPT, 200620–26	Low	Low	Low	Low	Low	Low	Low	High
Campbell et al., 199427	Unclear	Unclear	High	High	Low	Unclear	Low	Unclear
Collier et al., 198928	Unclear	Unclear	High	High	Unclear	Unclear	Low	High
DeFronzo et al., 199529	Unclear	Unclear	Unclear	Unclear	Unclear	Unclear	Low	High
Derosa et al., 200442	Unclear	Unclear	High	High	Low	Unclear	Low	Unclear
Hermann et al., 1991a30	Low	Unclear	High	High	Unclear	Unclear	Low	High
Hermann et al., 1991b31–34	Low	Low	Low	Low	Low	Low	High	High
Kamel et al., 199735	Unclear	Unclear	Unclear	Unclear	Unclear	Unclear	Low	Unclear
Lawrence et al., 200436	Unclear	Unclear	High	Low	Low	Unclear	Low	High
Tang et al., 200441	Unclear	Unclear	High	High	Unclear	Unclear	Low	Low
Tessier et al., 199937	Unclear	Unclear	High	High	Low	Unclear	Low	High
Tosi et al., 200338	Low	Low	Low	Low	Unclear	Low	Low	High
UKPDS 34, 19982,39,40	Low	Low	High	Low	Unclear	High	Low	High
Yamanouchi et al., 200543	Low	Low	High	High	Low	Unclear	Low	Unclear

Note: ADOPT = A Diabetes Outcome Progression Trial, UKPDS = United Kingdom Prospective Diabetes Study.
*The Cochrane risk-of-bias tool was used to assess the risk of bias for each study. Low risk = bias, if present, is unlikely to alter the results seriously, unclear risk = bias raises some doubt about the results, high risk = bias may alter the results seriously.¹⁰