SAMURAI: Sensitivity analysis of a meta-analysis with unpublished but registered analytical investigations (software)

Table 1 Default risk ratios (RR) assigned to unpublished studies with binary outcomes

Outlook	Depends on published studies?	If the outcome event is desired (higher.is.better = T)	If the outcome event is undesired (higher.is.better = F)
‘very positive’	No	3	0.33
‘positive’		2	0.5
‘no effect’		1	1
‘negative’		0.5	2
‘very negative’		0.33	3
‘very positive CL’	Yes, depends on CL of published studies	UCL _pub,1-α	LCL _pub,1-α
‘positive CL’		$0.5 ({\hat{RR}}_{pub} + UC L_{pub, 1 a})$	$0.5 ({\hat{RR}}_{pub} + LC L_{pub, 1 a})$
‘current effect’		${\hat{RR}}_{pub}$	${\hat{RR}}_{pub}$
‘negative CL’		$0.5 ({\hat{RR}}_{pub} + LC L_{pub, 1 - a})$	$0.5 ({\hat{RR}}_{pub} + UC L_{pub, 1 - a})$
‘very negative CL’		LCL _pub,1-α	UCL _pub,1-α

Note: ‘Positive’ outlooks favor the experimental treatment arm and ‘negative’ outlooks favor the control treatment arm.
RR: Risk ratio, or relative risk.
LCL: Lower confidence limit.
UCL: Upper confidence limit. ${\hat{RR}}_{pub}$ : The summary RR across all published studies.
LCL _pub,1-α,UCL _pub,1-α: The 1-α confidence interval of ${\hat{RR}}_{pub}$ .

ISSN: 2046-4053