The broad, long-term objectives of this proposal are to improve the outcome of patients with valvular heart disease and to expand the use of meta-analysis beyond the realm of randomized controlled trials (RCTs). The choice of prosthesis in valve surgery is a common and difficult clinical decision with large implications for patient outcomes as well as the cost of care. A meta-analysis of patient outcomes after valve surgery with different prostheses would provide valuable information for patients and physicians considering choice of prosthesis. A large body of data pertaining to this problem, however, is outside the realm of RCTs and involve longitudinal (time-dependent) data. Statistical techniques that go beyond those commonly used in meta-analysis are required. We propose to fill this gap in methodology by developing new approaches to longitudinal and non-RCT data and using these methods to complete a meta- analysis of an important clinical problem.
Our specific aims are: 1) Perform a meta-analysis of patient outcomes after valve replacement surgery. Outcomes of interest include perioperative death, long-term survival, complications, reoperation, functional status and prosthetic valve failure. Data sources for the meta-analysis include a small number of RCTs, some nonrandomized comparative trials and a large number of surgical case series. We will use well-established meta-analytic methods to assemble, grade and abstract the eligible literature. 2) Develop and apply new statistical methods to perform meta-analysis for non-RCT and longitudinal data. We plan to address both descriptive and comparative analyses of both single-point-in-time and longitudinal measures. Data from non-controlled case series reports and data with longitudinal outcome reports (e.g., survival curves) are of special interest because they are very common in the valve surgery literature and because analytic approaches have not yet been described. The data available for this clinical problem are similar to those available for many other medical technologies: few RCTs but many observational reports. Therefore, we anticipate that these new methods will have wide application beyond the issue of valve surgery and will be broadly generalizable to other research questions in health care. At completion, we expect to have reliable summary estimates of the patient outcomes after valve replacement surgery and new statistical approaches to meta-analysis that will be useful in many other applications.
