For AF study
The primary outcome is a composition of newly occurred ischemic stroke/transient ischemic attack (TIA) with positive neuroimaging or systemic arterial embolism (ICD-9 codes 434.x or 444.x [thromboembolic stroke or systemic embolism] or 435.x [transient ischemic attack]), and cardiovascular mortality during one-year follow-up.
Secondary outcomes include: cardiovascular mortality, newly occurred ischemic stroke, newly occurred transient ischemic attack, newly occurred hemorrhagic stroke (ICD-9 codes 430-432), and bleeding events of BARC type III, IV, and V.23 AF-related health utilization, measured by the occurrence of outpatient visits and hospitalizations with a diagnosis of AF, will also be recorded.
For non AF study
The primary outcome is a composition of newly occurred ischemic stroke/transient ischemic attack and cardiovascular mortality during one-year follow-up.
Secondary outcomes include postoperative AF (defined as newly diagnosed AF within 30 postoperative days [ICD-9 427.31; ICD-10 I48.0, I48.1, I48.2, I48.91]), cardiovascular mortality, newly occurred ischemic stroke, newly occurred transient ischemic attack, newly occurred hemorrhagic stroke (ICD-9 codes 430-432), bleeding events of BARC type III, IV, and V, and AF-associated health utilization.
Sample size calculation
The estimation of the primary outcome in the control group is based on reasonable assumptions about the patient risk and the possible types of antithrombotic therapy during follow-up (Table 2 and Figure 1).
For AF study
The primary outcome in AF group is estimated to be 7.6 per 100 person-year in non SLAAO group. Assuming 80% power and two-sided type I error of 0.05, we need to enroll 1016 patients in each group to detect a 40% relative risk reduction in the primary outcome in SLAAO group, accounting for 5% of patients’ loss during the one-year follow-up.
The scheme for estimation of the primary outcome for AF study is as follows (Table 2 and Figure 1):
Based on previous large registries,24 the proportion of patients receiving mechanical valve replacement is estimated to be 63.2%. We assume that all these patients take warfarin and adhere to standard warfarin medication during one year follow-up.
As for those receiving bioprosthesis or valve repair, four medication conditions are assumed: taking warfarin, aspirin, new anticoagulant, or no anticoagulant therapy. According to a extensive insurance database, we have reasonable estimates of the proportions of the four medication conditions in Chinese patients with AF.25
Based on the recent large trials, we estimate the newly occurred ischemic stroke/TIA in the control group taking warfarin will be 1.7 per 100 person year;26 taking aspirin will be 3.7 per 100 person year;27 taking new anticoagulants (dabigatran and the Factor 10a) will be 1.5 per 100 person year;26, 28 without any anticoagulant therapy will be 5.1 per 100 person year.27, 28
The ischemic stroke/TIA event rate in the control arm is estimated at 2.6 per 100 person-year. Because the postoperative mortality is 4.3-9.2 per 100 person-year based on studies from a large US administrative database and the cohort from the Society of Thoracic Surgeons Adult Cardiac Surgery Database,21, 22 we assume that cardiovascular mortality will be 5.0 per 100 person-year.
Thus, the overall event rate’s final estimates will be 7.6 per 100 person-year in the control arm without SLAAO.
For non AF study
Assuming 1059 patients in each group, we would detect a relative reduction rate of at least 40%, with a power of 80% and two-sided type I error of 0.05, in the primary outcomes with an estimated control event rate of 6.8 per 100 person-year.
The scheme for estimation of the primary outcome for AF study is as follows (Table 2 and Figure 1):
As aforementioned, we could easily estimate the annual stroke/TIA event rate in patients receiving mechanical valve replacement who will take warfarin for life long time.
As for those receiving bioprosthesis or valve repair, we first categorize the patients into three proportions: CHA2DS2-VASc score=2, CHA2DS2-VASc score=3, CHA2DS2-VASc score≥4. We have estimates about the three proportions from a recent large registry study.29 We also have estimates about the occurrence rate of postoperative AF (POAF) in the three categories.30, 31
For patients receiving bioprosthesis/valve repair and not developing POAF, we have good estimates of the stroke/TIA in each category.29
For patients receiving bioprosthesis/valve repair and developing POAF, we assume four medication conditions and have a reliable estimate of the four proportions: taking warfarin, aspirin, new anticoagulant, or no anticoagulant therapy.25 We also have estimates of the ischemic stroke/TIA event rate in the four medication conditions described in “For AF study (the prospective longitudinal study)” section.26-28
Thus, the ischemic stroke/TIA event rate in the control arm is estimated at 1.8 per 100 person year. We assume that cardiovascular mortality will be 5.0 per 100 person-year. Then we get the final estimates of the overall event rate of 6.8 per 100 person-year in the control arm without SLAAO.
Case report form (CRF) abstraction, follow-up, and data process.
Research staff from each site will scan all the patients’ medical charts in either the prospective longitudinal study or the randomized controlled trial, then transmitted the scanned copy to the coordinating center through the mail on encrypted, password-protected flash drives. The CRF will be quality-controlled, and the medical records will be de-identified by hiding all personal information in the records.
The CRFs include the patients’ baseline information (age, gender, and cardiac/non-cardiac history, et al.), invasive/non-invasive testing (ECG, echocardiography, chest X-ray, CT scans, angiography, et al.), laboratory results, in-hospital medications and surgical interventions, in-hospital complications, and discharge medications (Table 3). Trained abstractors will abstract this information under the supervision of trained quality control personnel. 10% of these records will be randomly selected for review by project managers to ensure adherence to the research protocol.
The detailed protocol for the follow-up was described elsewhere.32 Briefly, patients discharged alive are interviewed at the time point of discharge, postoperative 30 days, three months, six months, and one year. Face to face interview is the most preferred approach, but the telephone interview is also acceptable. The data will be stored at the coordinating center and protected in an encrypted and password-protected database.