Methods

The study cohort was drawn from a consecutive series of patients who had chronic severe nonischemic isolated MR and who were enrolled in our prospective study of the natural history of regurgitant valvular diseases (details of the study have been described ). Briefly, at study entry, patients undergo an upright ETT as well as other noninvasive objective cardiac tests. Testing is performed annually as part of routine clinical follow-up, although occasionally studies are not available if not ordered by the primary treating physician. Clinical characteristics are noted at entry and reassessed annually thereafter. Patients are excluded if, at entry, they have clinically evident coronary artery disease or greater than mild additional valve disease. Referral for mitral valve surgery is not mandated by protocol and is determined by the patient’s treating physician. This study protocol was approved by the Committee on Human Rights in Research of Weill Cornell Medical College (New York, NY) from its inception until September 2008 and has been approved by the Institutional Review Board of SUNY Downstate Medical Center (Brooklyn, NY) from October 2008 to the present.

For inclusion in this analysis, patients must have undergone mitral valve surgery (replacement or repair) for chronic, nonischemic, isolated, hemodynamically severe (3+ or 4+) MR (confirmed by preoperative cardiac catheterization or echocardiography ) without hemodynamic evidence of stenosis, and preoperative ETT with electrocardiography. Between February 1980 and June 2008, 96 patients from our natural history study with isolated, severe, nonischemic, chronic MR underwent mitral valve surgery. Of these, 18 were excluded because, at the time of operation, they had developed coronary artery disease or additional hemodynamically significant other valve disease; 33 additional patients were excluded because an ETT was not performed within 15 months before surgery (the cutoff for inclusion in this study). After these exclusions, 45 patients remained and comprised the cohort for the present analysis.

Clinical variables documented preoperatively included age, gender, NYHA-FC, MR etiology (inferred from echocardiography and historical data), history of atrial fibrillation, use of cardiac medications, and type of surgery (replacement or repair; Table 1 ). All 45 patients underwent symptom-limited upright ETT with electrocardiography, average 4.2 ± 4.4 (range 0–14.4) months before surgery. ETT was performed according to the modified Bruce protocol, including 3 minutes of exercise initially at grade 0 ^° at 1.7 miles per hour. Exercise duration was defined as the number of minutes on ETT, including time at grade 0 ^° ; also determined at ETT were chronotropic index, heart rate recovery, % maximum predicted heart rate achieved, blood pressures (rest and peak exercise) and heart rates (rest and peak exercise), according to standard methods previously described ( Table 2 ). Electrocardiographic response to exercise was considered “positive” only when ≥0.1 mV (1 mm) additional horizontal or downsloping ST-segment depression occurred during exercise compared with rest at 60 to 80 milliseconds after the J point; upsloping ST-segment depression and/or subthreshold ST depression were considered “negative.” Echocardiography (2-dimensional or Doppler) was performed at rest. LV systolic dysfunction was defined as an echocardiographic ejection fraction <50% or fractional shortening <25%.

Patients were followed approximately annually after surgery through February 2012. Follow-up included history and physical examination (by a study physician or, when this was not possible, by telephone calls to the patient, family, and patient’s primary care physician, supplemented by review of medical and vital records to determine vital status, including cause of death, and occurrence of valve reoperations). Cardiovascular deaths were defined as those due to heart failure, myocardial infarction, stroke, or those that were sudden. Postoperative NYHA-FC was assessed annually through 2 years. Average follow-up among nonreoperated survivors was 9.2 ± 4.3 years; 80% of patients were followed to death or to ≥5 years without reoperation.

Descriptive statistics are presented as mean ± SD (continuous variables) or number and percent (categorical variables). Student t tests, Pearson Chi-square tests or Fisher exact tests, as appropriate, were used to compare preoperative subgroup characteristics. Kaplan-Meier product-limit estimate curves were constructed and compared by log-rank tests to evaluate the univariable relation of preoperative exercise duration to postoperative death (all causes, cardiovascular) and for calculation of absolute and average annual risks; univariable Cox model analysis was used for calculation of hazard ratios and associated confidence intervals. Also tested by univariable survival analysis were commonly documented clinical variables ( Table 1 ), as well as other ETT descriptors and echocardiographic parameters ( Table 2 ). For the analysis of exercise tolerance, exercise duration (in minutes) by ETT was prospectively stratified according to statistical terciles: >11 minutes exercise time (upper tercile), 8 to 11 minutes (middle tercile), ≤7 minutes (lower tercile, equivalent to 1 minute of Stage 2 of the standard Bruce protocol, reflecting fairly low effort capacity); due to statistically indistinguishable mortality risks, the upper 2 terciles were combined, and the aggregated data were compared with those of the lowest tercile. Other noncategorical descriptors were stratified by previously identified high-risk cutpoints or according to statistical terciles when cutpoints had not been previously defined (with the “lowest risk” terciles aggregated). Because of the small number of endpoints, a full multivariable model containing all potential explanatory variables could not be constructed; thus, multivariable analyses were conducted pairwise (for exercise duration vs each of the other covariates) to adjust for the potentially confounding influence of preoperative variations in age, gender, use of cardiac medications, and peak exercise heart rates, as well as to explore the independent value of exercise duration versus other variables for prediction of postoperative death. Descriptors entered into the multivariable models were stratified according to the same cutpoints used for univariable analysis to simplify interpretation and because separate analysis of data (not shown) suggested that the corresponding hazards of many of the covariates did not obey the usual linearity assumptions of the Cox model. Follow-up time for all survival analyses was indexed to date of preoperative assessment and was censored at reoperation to mitigate the potential influence of reoperation on survival in patients who underwent repeat mitral valve surgery (4 patients). Fisher exact tests were performed to evaluate the association of preoperative exercise duration and other variables to symptom persistence ≤2 years after surgery among patients who were symptomatic before operation. The groupings used to stratify the predictor variables included in these analyses were identical to those used in the survival models. The criterion for statistical significance was a p value of <0.05 (2-sided test).