GRADE evidence profile for participants in randomised controlled trials (RCTs) compared with eligible non-participants who received similar treatment

Outcome	No of comparisons	Design	Study limitations	Consistency	Directness*	No of patients in RCT	No. of patients outside RCT	Relative effect (95% CI)	Overall quality
Dichotomous	73	Observational	No serious limitations†	Important inconsistencies‡	Direct	30 458	19 640	0.94 (0.87 to 1.02)	Very low
Continuous	18	Observational	No serious limitations†	No important inconsistencies	Direct	713	764	0.01 (-0.10 to 0.12)§	Low
Mortality	32	Observational	No serious limitations†	Important inconsistencies‡	Direct	26 406	16 609	1.00 (0.91 to 1.09)	Very low

GRADE Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004;328:14.

*Studies vary widely in terms of participants, interventions, and outcome measures; directness (applicability) of overall results to any specific trial is difficult to assess, particularly in light of unexplained heterogeneity of results.

†Although many studies were poorly controlled, differences in risk of bias did not seem to be related to differences in results, and overall results are consistent with results for studies without serious limitations.

‡Important, unexplained, statistically significant heterogeneity.

§Standardised mean difference (95% confidence interval).