| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Research and Development Office, Northport Veterans Affairs Medical Center, 79 Middleville Rd (151), Northport, NY 11768
b Division of Cardiac Research, Department of Veterans Affairs and, Department of Biostatistics and Informatics, University of Colorado Denver, Box B-119, 4200 E 9th Ave, Denver, CO 80262
(Email: annie.shroyer{at}va.gov; gary.grunwald{at}ucdenver.edu).
In this timely and clinically relevant article, MacKenzie and colleagues [1] have developed statistical models for predicting survival up to 8 years after coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI) based on 10 plus years (> 35,000 records) from the Northern New England Cardiovascular Disease Study Group (NNECDSG) database. They have quantified the major risk associations and made the resulting models and survival predictions available to investigators and clinicians.
MacKenzie and colleagues [1] elegantly applied sophisticated statistical methods to meet major analytical challenges, including time to event outcomes (Cox survival models), nonproportional hazards over long time periods (clever use of time partitions), nonlinearity of risk associations (splines), different associations of risk variables with postintervention survival at different times (separate models within time partitions), interactions between risk variables, and model validation and calibration. This provides a very comprehensive evaluation of both short- and long-term survival after CABG and PCI.
These NNECDSG results align with several Continuous Improvement in Cardiac Surgery Program (CICSP) database publications. For example, the tendency for cardiac risk factors to be important in the short term, but less so in the long term (see Table 2; [1]) is concordant with CICSP post-CABG findings [2]. The comprehensive nature of MacKenzie and colleagues' [1] analytic methods will aid in establishing patterns across adult cardiac databases.
One issue that arises in studying effects of processes of care on outcomes in observational studies is patient selection bias. Estimated survival probabilities include the likelihood of receiving a CABG or PCI, and can provide patients a realistic picture of their survival chances assuming their treatment was determined in a way similar to the original study. Thought is needed in comparing CABG and PCI since treatments were not randomized. Statistical methods for estimating treatment effects in observational studies are available, although sometimes controversial, and they remain an active area of statistical research [3, 4]. However, in the current setting, clinical factors that indicate, contraindicate, or preclude the use of one treatment over the other can be complex and not easily captured by available risk factors, particularly in the extremes of risk [5, 6]. Such clinical factors and decisions may also be hospital or clinician dependent.
Inherently, analytical challenges in this setting remain. MacKenzie and colleagues' [1] approaches provide a basis for several interesting future extensions. The use of survival analysis assumes there is stability of treatment approaches across the study period (i.e., the early and late study patients within treatment should have the same survival probability). Given the dynamic nature of advancements in PCI procedures (e.g., new devices, medications), it would be interesting to analytically incorporate changes in PCI-related procedural details over time. Another challenge is how to optimize patient-centered care using a combination of phased treatment approaches over time. For phased PCI and CABG performed during the same admission, a data-driven approach for the relative sequencing and timing of multiple procedures within the same admission, as well as across multiple admissions, has not yet been identified.
The excellent and comprehensive statistical prediction methods used by MacKenzie and colleagues [1] suggest their use in a broader context. The NNECDSG statistical methods seem to be widely applicable. Given regional variations in patient risk characteristics and outcomes, their prediction equations should be assessed for potential application in other systems. To evaluate the broader-based applicability of MacKenzie and colleagues' [1] equations and predictions beyond the New England region, a new "gold standard" de-identified adult cardiac surgery national registry should ideally be developed. If/when available, this national repository may facilitate assessment of the broader-based applicability for future cardiac surgical risk-modeling innovations.
 |
Acknowledgments |
|---|
 Top Acknowledgments References |
|---|
|
References |
|---|
|
TopAcknowledgmentsReferences |
|---|
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
