Table 1.

Clinical input parameters

	SAVR	Distribution	Source	TAVI	Distribution	Source
Early mortality (%)
After initial intervention	3.9*	Multivariate normal7	ACSD	5.4	Beta (α 65, β 1135)	[21]
After re-intervention	9.0*	Multivariate normal7	ACSD	8.63	Uniform (± 10%)	[41]
Early events (%)
Stroke	2.5*	Multivariate normal7	ACSD	2.9	Beta (α 58, β 1919)	[21]
Myocardial infarction	1.6*	Multivariate normal7	ACSD	1.0	Beta (α 20, β 1983)	[21]
Vascular complications	–	–		8.1	Beta (α 50, β 565)	[21]
Bleeding1	4.2	Beta (α 77, β 1761)	[8]	8.7	Beta (α 11, β 115)	[21]
Arrhythmias/atrial fibrillation	41.5*	Multivariate normal7	ACSD	11.0	Beta (α 31, β 249)	[21]
Pacemaker implantation (PI)	8.1	Beta (α 4, β 48)	[8]	12.2	Beta (α 85, β 610)	[21]
Renal failure/acute kidney injury	3.4*	Multivariate normal7	ACSD	4.5	Beta (α 10, β 215)	[21]
Prosthetic valve dysfunction2	–	–	Assumption	6.8	Beta (α 30, β 405)	[21]
Prosthetic valve thrombosis	–	–	Assumption	–	–	Assumption
Prosthetic valve endocarditis	–	–	Assumption	–	–	Assumption
Late events (%/year ± SD)
Stroke	0.77 ± 0.28	Lognormal	[8]	0.96 ± 0.104	Lognormal	[8, 42]
Probability of dying (%)	44.0	Beta (α 11, β 14)	[8]	44.0	Beta (α 11, β 14)	[8]
Bleeding	0.75 ± 0.16	Lognormal	[8]	0.95 ± 0.354	Lognormal	[8, 42]
Probability of dying (%)	39.1	Beta (α 18, β 28)	[8]	39.1	Beta (α 18, β 28)	[8]
Structural valve deterioration	Rate: 0.003 ± 0.001; Shape: 0.124 ± 0.024	Gompertz	[8]	Lognormal; mean log 2.711 ± 0.379; SD log 0.613 ± 0.335	Lognormal	[7]
Probability of dying (%)	17.0	Dirichlet6 (α1 18, α2 45, α3 41)	[43]	17.0	Dirichlet6 (α1 18, α2 45, α3 41)	[43]
Probability of re-intervention (%)	43.3		[8]	25.0		[7]
Probability TAVI	6.2	Uniform (6.1–6.3)	[5]	100	Assumption
Probability SAVR	93.8		[5]	0	Assumption
Probability conservative treatment	39.7			58.0	Assumption
Probability TAVI	61.7	Uniform (42.0–81.7)	[5]	0	Assumption
Probability medical treatment	38.3		[5]	100	Assumption
Nonstructural valve dysfunction	0.47 ± 0.27	Lognormal	[8]	–	Assumption
Probability of dying (%)	5.0	Dirichlet6 (α1 1, α2 10, α3 15)	[43]	–	–
Probability of re-intervention (%)	38.5		[8]	–
Prosthetic valve thrombosis	0.12 ± 0.09	Lognormal	[8]	0.245	Uniform (± 20%)	[44]
Probability of dying (%)	0.0	Dirichlet6 (α1 0, α2 2, α3 15)	[43]	0.0	Dirichlet6 (α1 0, α2 3, α3 23)	[43]
Probability of re-intervention (%)	0.12		[45]	0.12		[46]
Prosthetic valve endocarditis	0.57 ± 0.08	Lognormal	[8]	0.54 ± 0.10	Lognormal	[21]
Probability of dying (%)	34.0	Dirichlet6 (α1 26, α2 37, α3 13)	[43]	34.0	Dirichlet6 (α1 26, α2 37, α3 13)	[43]
Probability of re-intervention (%)	49.0		[8]	49.0		[8]
Hazard ratio excess mortality	0.86	Uniform (± 10%)	[8]	1.40	Uniform (± 10%)	This study

*Mean (95% CI) in the Adult Cardiac Surgery Database (ACSD). Risk in the patient-level simulation model dependent on patient and intervention characteristics using logistic regression formula. “-“Not reported in any of the studies, therefore assumed not to occur

¹Definition of bleeding is reexploration for bleeding after SAVR and major bleedings after TAVI

²Paravalvular leak after TAVI

³Hazard ratio of 1.6 applied to early mortality risk of initial intervention

⁴Hazard ratio of SAVR patients compared to the general population applied to occurrence in age and sex matched general population for the TAVI population

⁵Blackstone & Kirklin have shown that valve thrombosis mainly occurs during the first year after surgical mechanical aortic valve implantation and deteriorates to almost zero after six years [47]. The higher occurrence in the early phase may be caused by suboptimal anticoagulation treatment in the first post-intervention period. Since, the mean follow-up of the Bern TAVI Registry was only one year, it is likely that the occurrence rate of valve thrombosis after TAVI found in this study will not remain constant but will reduce over time. Therefore, we recalculated the linearized occurrence rate of 0.69%/patient-year, assuming that it will be zero from year 7 onwards

⁶Dirichlet distribution parameters: α¹ = number of deaths, α² = number of re-interventions, α³ = number of other treatment

⁷Multivariate normal distribution: coefficients of the regression model are randomly drawn from a multivariate normal distribution based on coefficients and variance–covariance matrix