Two-stage Bayesian hierarchical modeling for blinded and unblinded safety monitoring in randomized clinical trials

Table 7 The probability of flagged trials for the AEs under three AEs with flat effect relationships and same effect for the rest with safety issues for scenario IV

Example AEs (True Proportion; Expected Event Rate)	Blinded: Proportion with signal for early termination Unblinded: Termination Proportion with confirmed by unblinded data	(P_crit1, P_crit2)
Example AEs (True Proportion; Expected Event Rate)		(0.9, 0.7)	(0.9, 0.8)	(0.9, 0.9)
Pneumothorax Induced by HBO therapy^a (π = 3%; π_M = 2%)	Blinded	0.25	0.26	0.26
	Unblinded	0.74	0.57	0.28
	Overall Rate	0.19	0.15	0.07
Signs of Pulmonary Dysfunction^a (π = 31%; π_M = 25%)	Blinded	0.53	0.53	0.53
	Unblinded	0.74	0.60	0.39
	Overall Rate	0.39	0.31	0.21
Pneumonia^a (π = 48%; π_M = 40%)	Blinded	0.56	0.57	0.58
	Unblinded	0.75	0.61	0.41
	Overall Rate	0.43	0.35	0.24
Critical decreased CPP^a (< 60 mmHg) (π = 80%; π_M = 75%)	Blinded	0.50	0.49	0.51
	Unblinded	0.66	0.51	0.30
	Overall Rate	0.33	0.25	0.15
Critical hypotension^a (MAP< 70 mmHg) (π = 80%; π_M = 75%)	Blinded	0.49	0.50	0.51
	Unblinded	0.66	0.51	0.29
	Overall Rate	0.32	0.26	0.15
Seizures during HBO treatment^a (π = 1%; π_M = 1%)	Blinded	0.21	0.22	0.22
	Unblinded	0.74	0.56	0.28
	Overall Rate	0.16	0.12	0.06
Hypercarbia during transportation^a (PaCO2 > 45 mmHg) (π = 13%; π_M = 10%)	Blinded	0.39	0.39	0.38
	Unblinded	0.67	0.50	0.28
	Overall Rate	0.26	0.19	0.11

ISSN: 1471-2288