Table 2: Annual transition probabilities.

Model item Mean annual probability Sources/underlying assumptions

Incidence of AF in ICD patients20.94%Own calculation based on publications by Safak et al. 2013 [12], Bunch et al. 2009 [24], Mittal et al. 2008 [25] and Healey et al. 2012 [3]

Proportion of AF detected (Lumax VR-T DX system)98.00%Derived from Seidl et al. 1998 [14] Recording of atrial tachyarrhythmias based on high atrial rate detection.

Proportion of AF detected (conventional ICD)15.00%AF episodes recording based on single-chamber ICD detection criteria. Confirmation of AF by 24-hour Holter monitoring. Own calculation based on publications by Friedmann et al. 2006 [26], Moss et al. 2012 [27] and Charitos et al. 2012 [28]

Stroke incidence in untreated AF-patientsBased on CHA2DS2VASc, for base case: 3.96%Directly derived according to the CHA2DS2-VASc-Score. For the base-case scenario, a weighted incidence was calculated according to the distribution of CHA2DS2-VASc-Scores found in German Heart Center Munich

Stroke incidence in no AF-patientsfor base case:
0.82%
According to Wolf et al. 1991 [29], the risk of stroke is 4.8 fold increased if a patient is suffering from AF. In reversing circuit, the stroke incidence in patients without AF is the stroke incidence in untreated AF patients / 4.8

Stroke incidence in AF-patients receiving OACfor base case:
1.38%
According to Hart et al. 2007 [7], anticoagulation using Warfarin is associated with a mean stroke risk reduction of 64% (49% to 74%). Stroke incidence in AF patients receiving Warfarin was calculated as follows: Stroke incidence in untreated AF patients (1-0.64). Since NOACs are slightly more effective in reducing strokes than Warfarin (Patel at al. 2011 [20]), we considered a stroke risk reduction of 66% for NOACs compared to untreated AF patients. The stroke incidence used in the model for OAC was calculated as a weighted mean according to the distribution of NOACs and Warfarin (52.1% vs. 47.9%).

Stroke mortality in no AF-patients34.00%Directly derived from Lin et al. 1996 [30] where 41 of 120 non AF patients died within one year

Stroke mortality in untreated AF-patients63.00%Directly derived from Lin et al. 1996 [30] where 19 of 30 AF patients died within one year

Stroke mortality in AF-patients receiving OAC42.00%According to Lip et. al. 1996 [19] where anticoagulation with warfarin reduced mortality by a third. Stroke mortality in AF patients receiving anticoagulation was calculated as follows: Stroke mortality in untreated AF patients (1-1/3)

Mortality in no AF-patients, no stroke (background mortality)6.00%Directly derived from van Welsenes et al. 2011 [31]. The cumulative incidence for all-cause mortality was 6% at year 1.

Mortality in untreated AF-patients, no stroke11.10%Own calculation based on Stewart et al. 2002 [32] who reported a mortality increase of RR=2.2 in women and 1.5 in men. We assumed a mean mortality RR for both sexes of 1.85. This was multiplied with background mortality of 6%

Mortality in AF-patients receiving OAC, no stroke7.40%According to Lip et al. 1996 [19] where anticoagulation with Warfarin reduced mortality by one third. Mortality in AF patients receiving anticoagulation was calculated as follows: Mortality in untreated AF patients (1-1/3)

Risk for major bleeding under OAC3.50%Directly derived from ROCKET AF study (Patel et al. 2011 [20]) where the event rate of major bleeding was 3.4 per 100 patient-years for patients with nonvalvular AF and treatment with Warfarin. The corresponding bleeding rate under Rivaroxaban was reported as 3.6. The risk for major bleeding used in the model was calculated as a weighted mean according to the distribution of NOACs and Warfarin (52.1% vs. 47.9%).

Risk for major bleeding without OAC2.95%Own calculation based on Go et al. 2003 [22] where the crude rate of major hemorrhage was 1.52 events (1.28 events) per 100 patient-years in a real world patient population with nonvalvular AF receiving anticoagulation (resp. not receiving anticoagulation). Only events that did lead to hospitalization were analyzed in this study. This may have been associated with a risk of underreporting of events. Therefore, we multiplied the estimated reduction of major bleeding events (1.28/1.52) without OAC with the rate of major bleeding events under OAC used in the present model.

AF: atrial fibrillation; ICD: implantable cardioverter-defibrillator; NOACs: new oral anticoagulants; OAC: oral anticoagulation; RR: relative risk.