Author + information
- Robert H. Jones, MD, FACC⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Robert H. Jones, Duke Clinical Research Institute, P.O. Box 17969, Durham, North Carolina 27715.
Truth reflected in 2 common-but-conflicting aphorisms must coexist in dynamic tension to produce favorable outcomes in cardiac surgical patients: “fix everything while you are there” must be balanced with “perfect is the enemy of good” to do all but not more than is needed for the welfare of the patient. Fixing a leaking heart valve while one is already performing coronary artery bypass grafting (CABG) does not seem like excessive striving for perfection. However, the article by Mihaljevic et al. (1) in this issue of the Journalreports mitral valve annuloplasty produced no noticeable benefit for 54 of 290 patients undergoing mitral valve annuloplasty concomitantly with CABG when compared with 54 patients propensity matched for similar baseline characteristics from the concurrent 100-patient cohort who received only CABG. Do conclusions from these 108 of the 390 patients with 3+ and 4+ mitral regurgitation identified on an initial echocardiogram now forever end the long-standing debate about the value of adding mitral valve annuloplasty to CABG in patients with functional ischemic mitral regurgitation? Could analysis of only 38 deaths occurring over the course of 12 years in these 54 propensity-matched pairs of patients possibly justify the definitive title of this editorial? Can leaky valves possibly be safely left in the hearts of patients? The answers are yes and no to these questions. Implications of the two major author-defined limitations must be considered before accepting their conclusion as definitive.
Limitation of an Observational Study
Experienced authors know that the publication of an observational study in any major journal requires an explicit confession that the study was not randomized. Neither manuscript reviewers nor journal editors appear to seriously expect the randomized study to occur in their lifetimes. Therefore, only a simple statement of need appears to maintain the scientific status of the journal while not encumbering authors with unnecessary work to propose the design of a randomized trial suggested by the results of their observational study. Results of this observational study would support proposing a randomized clinical trial addressing the hypothesis that “not adding mitral valve annuloplasty to CABG increases the 6-year survival of patients with ischemic cardiomyopathy and 3+ or 4+ mitral regurgitation.” A feasibility estimate using the 6% absolute, but statistically insignificant, 6-year mortality advantage of not adding mitral valve annuloplasty to CABG suggests that at least 1,200 patients would need to be randomized and followed for a mean of 6 years to demonstrate a 6% mortality difference with a 90% statistical power. Neither the National Institutes of Health nor purveyors of mitral annuloplasty rings could reasonably be expected to provide the $30 to $50 million required to complete this study. An alternate hypothesis likely to be more attractive to medical device companies would be “adding mitral valve annuloplasty to CABG reduces short-term mortality.” However, the statistically insignificant (Table E1 in Mihaljevic et al. ) 5% excess death rate in the CABG-only population at 1 year (Fig. 1B in Mihaljevic et al. ) could be readily explained away by 5% of CABG-only patients for whom hemodynamic instability was the reason stated for no mitral valve annuloplasty. The title of this editorial comment states the most rational hypothesis to advocate for a more definitive clinical study. In a perfect world in which cardiovascular specialists were passionate about only offering therapy of proven benefit to their patients and Centers for Medicare & Medicaid Services was empowered by Congress to withhold reimbursement for unproven therapies except in the context of a randomized clinical trial, a prospective randomized trial to test a no-benefit hypothesis could easily be performed using cost savings generated from funds now expended for mitral annuloplasty procedures of questionable benefit. Given this unlikely scenario to fund a randomized trial of equivalence, the Mihaljevic et al. paper (1), interpreted in the context of 3 other observational studies on this topic (2–4) that failed to show benefit for adding mitral valve surgery to CABG in ischemic cardiomyopathy patients with functional mitral regurgitation, answers “yes” to the question as to whether “leaky valves can possibly be safely left in the hearts of patients.”
Limitation of Analysis by Operation Received
If cardiac surgeons still posted their own operations, the 3 different titles posted to definitively describe the specific operation intended for each patient in the 390-patient cohort would have been either CABG, CABG with mitral valve annuloplasty, or CABG with possible mitral valve annuloplasty. The posting of a possible mitral valve annuloplasty would more accurately indicate the cardiac surgeon to be at equipoise about addition of mitral valve annuloplasty to CABG in an individual patient than would propensity matching patients using baseline characteristics. A change in equipoise of the surgeon during operation might change any of the 3 operative titles into a decision for or against mitral surgery. The institution of CABG typically precedes mitral valve surgery so that good or bad events occurring early in the operation might weigh for or against additional surgery when the CABG is complete. Although retention of the option to exercise surgical judgment and change an intended operation is ethical and necessary to optimize operative results, confounding of reported outcomes can only be fully understood if operation-performed outcomes can be compared with operation-intended outcomes. Moreover, this analysis also must include all patients for whom a different operation was performed other than the one intended. Defining surgical cohorts retrospectively by operation performed permits bias intentionally or unintentionally masking as surgical judgment to exclude patients from cohorts by changing the name of the operation performed. For example, a patient undergoing a CABG with mitral valve annuloplasty that failed intraoperatively so as to require the mitral valve to be replaced would have been excluded in the Mihaljevic et al. (1) study by naming the operation a mitral valve replacement. Patients intended for CABG with or without mitral valve annuloplasty for whom an SVR operation was added when the patient failed to wean from cardiopulmonary bypass would be excluded by surgical ventricular reconstruction in the operation name. Without availability of a clear preoperative statement of operation intended, all clinical studies must define subgroups retrospectively by the operation received. Thereafter, an inventory can be taken of reasons stated for the operation performed. The dominant reason patients reported by Mihaljevic et al. (1) did not receive a mitral valve annuloplasty was that mitral regurgitation appeared less severe to the surgeon at operation than the ≥3+ mitral regurgitation reported on preoperative echocardiography. The 5 patients undergoing CABG who had hemodynamic instability noted as the reason a mitral valve annuloplasty was not performed must have had a long and difficult CABG that made the additional ischemic time required to complete the mitral surgery appear to be of questionable value. The constraint to use postprocedure names to define the cohort whose outcome information will be used to discuss the potential benefit of an operation with the same name with a preoperative patient introduces complex confounding of unknown magnitude and direction that cannot be fully neutralized by statistical modeling. Therefore, if the answer to the question “can leaky heart valves possibly be safely left alone” is to be answered with certainty, the answer must be no.
Although only required by study design in randomized clinical studies, preoperative intention to treat can be recorded equally easily for observational clinical studies. An ideal design for future observational or randomized clinical studies would begin with preoperative identification and obtaining consent for participation from potential patients who are made aware of their options for mitral valve surgery during CABG. At the point during the operation, when the responsible physician must decide whether consent given justifies study participation, the state of operating surgeon equipoise would define the operation choices compatible with the preoperative consent obtained. For the mitral annuloplasty question, this decision time would occur with patients on cardiopulmonary bypass after completion of the CABG and immediately before making the left atrial incision to begin the mitral valve repair. If the patient had previously consented to randomized selection of options that now coincide with surgeon equipoise, the specific operation intended would be defined by randomization. If the patient consented to only participate in a registry or observational study, the rationale for selection of a specific operation would be documented as the intention to treat, thereby minimizing need for subsequent crossover based on events that might change equipoise for a specific procedure. Insightful use of existing national cardiovascular databases supplemented with a web-based method to trap intention to treat intraoperatively now makes it logistically feasible to devise prospective clinical studies that could be inexpensively conducted on large numbers of patients. Until widespread demand for true evidence stimulates cardiovascular specialists to participate in gathering definitive evidence, we are left with imperfect but better evidence than we had heretofore. Cardiovascular specialists who have previously believed all leaky valves should be fixed now should downgrade their certainty to conclude that some valves probably should be fixed in certain patients who still remain difficult to identify before the operation begins.
The author is the principal investigator of the National Institutes of Health-sponsored Surgical Treatment for Ischemic Heart (STICH) Failure Trial.
↵⁎ Editorials published in the Journal of the American College of Cardiologyreflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
- American College of Cardiology Foundation
- Mihaljevic T.,
- Lam B.-K.,
- Rajeswaran J.,
- et al.
- Wu A.H.,
- Aaronson K.D.,
- Bolling S.F.,
- Pagani F.D.,
- Welch K.,
- Koelling T.M.