A, Unadjusted and cumulative Kaplan-Meier estimates of survival within 30 days after AVR surgery. B, Unadjusted and cumulative Kaplan-Meier estimates of survival within 30 days after hospital discharge.
eTable 1.CPT-4 and ICD-D-CM codes for CABG subgrouping
eTable 2. Complete demographic and clinical characteristics of Medicare patients undergoing AVR surgery (2006-2011) before and after propensity score matching
eTable 3. Risk for death following AVR among nonusers of warfarin in mixed-effect analyses
eTable 4. Estimated risk for death in the propensity score analyses
eTable 5. Model with interaction term between mechanical valve and female patients
eTable 6. Models in sex-specific subgroups
eFigure. Cohort selection flow chart
Customize your JAMA Network experience by selecting one or more topics from the list below.
Du D, McKean S, Kelman JA, et al. Early Mortality After Aortic Valve Replacement With Mechanical Prosthetic vs Bioprosthetic Valves Among Medicare Beneficiaries: A Population-Based Cohort Study. JAMA Intern Med. 2014;174(11):1788–1795. doi:10.1001/jamainternmed.2014.4300
Early mortality for patients who undergo aortic valve replacement (AVR) may differ between mechanical and biological prosthetic (hereinafter referred to as bioprosthetic) valves. Clinical trials may have difficulty addressing this issue owing to limited sample sizes and low mortality rates.
To compare early mortality after AVR between the recipients of mechanical and bioprosthetic aortic valves.
Design, Setting, and Participants
A retrospective analysis of patients 65 years or older in the Medicare databases who underwent AVR from July 1, 2006, through December 31, 2011. In the mixed-effects models adjusting for physician and hospital random effects, we estimated odds ratios (OR) of early mortality to compare mechanical vs bioprosthetic valves.
Mechanical or bioprostheticaortic valve replacement.
Main Outcomes and Measures
Early mortality was measured as death on the date of surgery, death within 1 to 30 or 31 to 365 days after the date of surgery, death within 30 days after the date of hospital discharge, and operative mortality (death within 30 days after surgery or at discharge, whichever is longer).
Of the 66 453 Medicare beneficiaries who met inclusion criteria, 19 190 (28.88%) received a mechanical valve and 47 263 (71.12%) received a bioprosthetic valve. The risk for death on the date of surgery was 60% higher for recipients of mechanical valves than recipients of bioprosthetic valves (OR, 1.61 [95% CI, 1.27-2.04; P < .001]; risk ratio [RR], 1.60). The risk difference decreased to 16% during the 30 days after the date of surgery (OR, 1.18 [95% CI, 1.09-1.28; P < .001]; RR, 1.16). We found no differences within 31 to 365 days after the date of surgery and within the 30 days after discharge. The risk for operative mortality was 19% higher for recipients of mechanical compared with bioprosthetic valves (OR, 1.21 [95% CI, 1.13-1.30; P < .001]; RR, 1.19). The number needed to treat with mechanical valves to observe 1 additional death on the surgery date was 290; to observe 1 additional death within 30 days of surgery, 121. Consistent findings were observed in subgroup analyses of patients who underwent concurrent AVR and coronary artery bypass graft, but not in the subgroup undergoing isolated AVR.
Conclusions and Relevance
In this cohort analysis of Medicare beneficiaries, use of mechanical aortic valves was associated with a higher risk for death on the date of surgery and within the 30 days after surgery compared with bioprosthetic aortic valves among patients who underwent concurrent AVR and coronary artery bypass graft but not isolated AVR.
Aortic valve replacement (AVR) is indicated for most patients with symptomatic aortic valve disease. Mechanical prosthetic and biological prosthetic (hereinafter referred to as bioprosthetic) valves are available as treatment choices. Mechanical valves have excellent durability but impose lifelong anticoagulant therapy on recipients.1 In contrast, bioprosthetic valves do not require continuous anticoagulant therapy, but patients are at a higher risk for reoperation owing to structural valve deterioration. According to the Agency for Healthcare Research and Quality,2 64 262 AVR procedures were performed in the United States in 2010 alone; of these, nearly 70% were conducted among patients 65 years or older. The current American Heart Association guidelines3 recommend bioprosthetic valves for patients 65 years or older based on the consensus opinion of experts.
Since the first AVR was reported in 1960,4 only 2 randomized controlled trials5,6 have been conducted to compare long-term outcomes in patients receiving mechanical prosthetic or bioprosthetic valves. Owing to their limited sample sizes, evaluation of early mortality in these clinical trials was impossible. In addition, most participants in these trials were younger than 65 years. As the population has aged, the number of AVR procedures performed among elderly Americans has increased.7 However, very limited empirical evidence is available for the selection of the appropriate prosthetic valves for patients 65 years or older.
In the present study, we compared early mortality after AVR between recipients of mechanical and bioprosthetic aortic valves using fee-for-service Medicare claims from July 1, 2006, through December 31, 2011. We focused on Medicare beneficiaries 65 years or older.
Our study was deemed exempt from review by the Research in Human Subjects Committee of the US Food and Drug Administration. The study cohort was composed of Medicare beneficiaries from July 1, 2006, through December 31, 2011. Data as early as January 1, 2006 were used to review patients’ records of diagnoses and use of medical care services. Medicare is a federal health insurance program in the United States that provides coverage for nearly 50 million Americans, including virtually all people 65 years or older and some younger adults with permanent disabilities or end-stage renal disease.8
Beneficiaries who were continuously enrolled in Medicare Parts A, B, and D for at least 6 months before their AVR procedures were eligible for the study. Patients in Medicare Part C were not included because limited claims data were available for these beneficiaries. Beneficiaries younger than 65 years or those originally enrolled in Medicare when they were younger than 65 years were excluded from the analysis because their Medicare entitlement is generally based on various permanent disabilities or end-stage renal disease, and therefore they are not representative of a more general Medicare population.
Patients who received mechanical or bioprosthetic aortic valves were identified by procedure codes 35.22 and 35.21, respectively, from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). To minimize bias, we excluded patients undergoing any open heart surgery within 6 months before their index surgery.
The study outcome was all-cause early mortality after AVR. Death was determined using the Social Security Death Index. Early mortality was evaluated for death on the date of AVR surgery, within 1 to 30 days or within 31 to 365 days after the date of AVR surgery, or within 1 to 30 days after the date of hospital discharge (not including death on the date of discharge), and operative mortality was defined as death during the procedure hospital stay or 30 days after surgery, whichever is longer.9 In addition, we reported unadjusted mortality rates within 1 to 7, 8 to 30, 31 to 60, and 61 to 90 days after AVR surgery.
Demographic variables were obtained from the common working file from the Centers for Medicare & Medicaid Services. Age was used as a continuous variable. Race was classified as white, black, or other. Socioeconomic status was assessed through the Low Income Subsidy indicator, which is based on patient income and resources.10 Patients were grouped into Low Income Subsidy recipients and nonrecipients. Geographic locations were grouped into North, Midwest, South, West, or other. Calendar years were treated as indicator variables for the year in which the surgery was performed. Surgical priority was evaluated by admission status as elective, urgent, emergency, or other.
Cardiovascular procedures concurrent with AVR involve complex health conditions and prolonged operation time and therefore may have higher rates of adverse outcomes. Coronary artery bypass graft (CABG), often performed concurrently during AVR procedures, was separated from other concurrent open heart surgical procedures. We further grouped CABGs into procedures involving 1 or 2 arteries, 3 or more arteries, the internal mammary coronary artery, and other coronary arteries (eTable 1 in the Supplement describes the coding). Concurrent CABG and other concurrent open heart surgical procedures were identified using ICD-9-CM procedure codes and were expressed as indicator variables based on the presence or the absence of the procedure. Other conditions that might affect the selection of valve type and surgery outcomes include the use of warfarin, history of hemorrhage, and the failure of previously implanted heart valves.3 Patients were considered warfarin users if they had a prescription for at least 31 days or a minimum of 3 prothrombin time and international normalized ratio (PT/INR) laboratory testing claims or 1 home PT/INR testing claim during the 6 months before the AVR procedure. Warfarin use was identified by National Drug Codes. The PT/INR tests were identified with Healthcare Common Procedure Coding System codes. History of hemorrhage and the failure of previously implanted heart valves were identified with ICD-9-CM codes.
Overall systemic health status was determined by the Elixhauser comorbidity index.11 Because the causes of 2 specific Elixhauser conditions, valvular diseases and cardiac arrhythmias, could affect the choice of prosthetic valve type and postoperative outcomes, valvular diseases were subdivided into endocarditis, rheumatic disease, congenital disease, and previous valve replacement with other procedures such as homograft replacement or Ross procedure, any other aortic valvular diseases, and no records of previous valvular diseases. Cardiac arrhythmias were regrouped into atrial fibrillation, supraventricular arrhythmias, atrioventricular block, ventricular fibrillation, and other cardiac arrhythmias. All comorbidities were identified as indicator variables based on the presence or the absence of the condition in inpatient, outpatient, and carrier claims during the 6 months before admission for the AVR procedure. Disease conditions were identified using ICD-9-CM codes.11 In addition, we controlled for the use of immunosuppressants, other anticoagulants, and cardiovascular agents. Claims for these drugs were identified with National Drug Codes. Patients were considered to use the drugs if they were prescribed a drug for more than 30 days during the 6 months before the AVR surgery.
Baseline patient characteristics were compared between recipients of mechanical and bioprosthetic aortic valves using the standardized mean difference.12 Kaplan-Meier curves were plotted to compare probabilities of post-AVR survival between recipients of mechanical and bioprosthetic valves for 0 to 30 days after the surgery and 1 to 30 days after the date of hospital discharge.13
Logistic regression modeling was used to examine the association of prosthetic valve type and study outcomes, adjusting for physician and hospital random effects with consideration of cross-classification and the covariates listed above.14 We calculated the risk ratio (RR) and the number needed to treat (NNT).15 All tests were 2 tailed and a P value of less than .05 was considered statistically significant. Most statistical analyses were conducted using the same commercially available software package (SAS, version 9.2; SAS Institute). The mixed-effect modeling with cross-classification was performed in R using the lme4 package.16
Subgroup analyses were performed in more restricted and homogeneous subgroups. Because other concurrent open heart surgical procedures could place patients at higher risk than the isolated AVRs, and because more than 45% of our study population underwent concurrent CABG, mixed-effect analyses were conducted for the subgroups undergoing isolated AVR and concurrent AVR and CABG. Lifelong anticoagulation is one of the major confounding factors of the study. Patients may prefer mechanical valves if they were undergoing long-term anticoagulation for existing indications, such as atrial fibrillation or venous thromboembolism, or after myocardial infarction.3,17 Inappropriately managed anticoagulation therapy could also increase the patient’s risk for stroke or bleeding. The risk for death due to the existing indications for anticoagulation and due to complications associated with anticoagulation could channel through mechanical valve assignment and confound the study results. Therefore, analyses were also performed in a subgroup not receiving warfarin. To assess whether the association between valve type and the risk for death on the date of surgery differed by sex, we tested the interaction between valve type and sex in the overall population and performed sex-specific subgroup analyses.
Because of differences in key baseline characteristics between mechanical and bioprosthetic valve recipients (eTable 2 in the Supplement), we also evaluated the study outcomes using a propensity score–matched method. The propensity score is the probability of a patient receiving a mechanical valve given a set of measured baseline covariates.18,19 Scores were estimated using a nonparsimonious multivariable logistic regression model in which the dependent variable is the selection of a mechanical aortic valve, and covariates are the potential confounders, baseline characteristics, and physician and hospital AVR volumes.20,21 Study cohorts were matched using a 1-to-2 (mechanical to bioprosthetic) greedy matching protocol without replacement and with a caliper width of 0.2 of the standard deviation of the logit propensity score.22,23 The standardized mean difference assessed the balance of covariates between cohorts before and after propensity score matching. As a rule of thumb, covariates with standardized mean differences of less than 0.10 were considered balanced between treatment groups.24,25 We then performed logistic regressions with physician and hospital random effects to estimate the ORs for death in recipients of mechanical vs bioprosthetic valves in the propensity score–matched cohorts.
We found 97 038 US Medicare beneficiaries who were enrolled in Parts A, B, and D and not C on the date of the AVR surgery from July 1, 2006, through December 31, 2011. Of these patients, 66 453 met the inclusion criteria; 19 190 (28.88%) received a mechanical valve and 47 263 (71.12%) received a bioprosthetic valve (eFigure in the Supplement). Table 1 shows the distribution of selected demographic and clinical characteristics of these patients. eTable 2 in the Supplement contains the complete list of demographic and clinical characteristics of all patients.
Table 2 shows early mortality rates after AVR and unadjusted ORs for death in recipients of mechanical compared with bioprosthetic valves. The death rates on the date of surgery were 1.04% and 0.57% for recipients of mechanical and bioprosthetic valves, respectively. The unadjusted OR for death between recipients of mechanical and bioprosthetic valves was highest on the date of surgery (unadjusted OR, 1.82 [95% CI, 1.52-2.19; P < .001]). The mortality rate for the first 30 days after the date of surgery was statistically higher among mechanical valve recipients than bioprosthetic valve recipients (6.09% vs 4.91%, respectively; unadjusted OR, 1.26 [95% CI, 1.17-1.35; P < 001]) but was not statistically significant for the first 30 days after the date of hospital discharge (2.56% for mechanical valves vs 2.38% for bioprosthetic valves; unadjusted OR, 1.08 [95% CI, 0.97-1.21; P = .18]) and within 31 to 365 days after the date of surgery (8.88% for mechanical valves vs 8.45% for bioprosthetic valves; unadjusted OR, 1.06 [95% CI, 0.99-1.12; P = .09]). The Figure illustrates the unadjusted survival curves comparing mechanical valves with bioprosthetic valves within the first 30 days after AVR surgery and within the first 30 days after the hospital discharge date.
In the mixed-effect analyses with physician and hospital random effects, patients who received mechanical valves had higher estimated rates of death on the date of surgery and within the short period after the date of surgery compared with those who received bioprosthetic valves (Table 3). Compared with bioprosthetic valves, mechanical valves were associated with a significantly greater risk for death on the date of surgery (adjusted OR, 1.61 [95% CI, 1.27-2.04; P < .001]; RR, 1.60; NNT, 290) and within the first 30 days after the date of surgery (adjusted OR, 1.18 [95% CI, 1.09-1.28; P < .001]; RR, 1.16; NNT, 121) and a significantly greater risk for operative mortality (adjusted OR, 1.21 [95% CI, 1.13-1.30; P < .001]; RR, 1.19; NNT, 83). In contrast, mechanical valves were not associated with an increased risk for death within 31 to 365 days after the date of surgery (adjusted OR, 1.03 [95% CI, 0.97-1.11; P = .32]) and within the first 30 days after the date of hospital discharge when compared with bioprosthetic valves (adjusted OR, 1.04 [95% CI, 0.93-1.17; P = .50]).
Results from the subgroup undergoing concurrent AVR and CABG were similar to results of the primary analyses. Compared with recipients of bioprosthetic valves, recipients of mechanical aortic valves had a significantly increased risk for death on the date of surgery (adjusted OR, 2.06 [95% CI, 1.08-3.93; P = .03]; RR, 2.06; NNT, 202) and within 30 days after the date of surgery (adjusted OR, 1.19 [95% CI, 1.04-1.38; P = .01]; RR, 1.18; NNT, 116) and had a significantly higher operative mortality (adjusted OR, 1.26 [95% CI, 1.11-1.43; P < .001]; RR, 1.23; NNT, 72). No association was observed within 31 to 365 days after the date of surgery (adjusted OR, 1.06 [95% CI, 0.95-1.19; P = .28]) and within the 30 days after the date of hospital discharge (adjusted OR, 0.92 [95% CI, 0.75-1.13; P = .45]). Results from the subgroup not receiving warfarin (eTable 3 in the Supplement) and propensity score–matched analyses (eTable 4 in the Supplement) yielded results consistent with the primary analyses. The association between prosthetic valve type and the risk for death on the date of surgery did not differ significantly by sex (the coefficient for the interaction between female sex and mechanical valve, 0.24 [P = .31]) (eTable 5 in the Supplement). The risk estimates comparing mechanical vs bioprosthetic valves on the date of surgery were almost identical in male (unadjusted mortality rate, 0.55%; adjusted OR, 1.46 [95% CI, 0.88-2.41]) and female (unadjusted mortality rate, 0.86%; adjusted OR, 1.45 [95% CI, 1.02-2.07]) subgroups (eTable 6 in the Supplement).
When compared with bioprosthetic aortic valves, mechanical aortic valves were associated with an increased risk for early death after concurrent AVR and CABG surgery in Medicare patients 65 years and older. The greatest risk existed on the date of surgery, when the adjusted RR for the mechanical valve cohort vs the bioprosthetic valve cohort was 1.60. A similar but less pronounced association also existed during the 30 days after AVR surgery, when the mechanical valve cohort experienced a 16% increased risk for death. Only when we measured the association during the 30 days after hospital discharge and within 31 to 365 days after the date of surgery did we see this risk disappear. These results suggested that a higher risk for death was associated with mechanical valves on the date of surgery and the period immediately after and that patients in both cohorts faced a similar risk for death during the short period after hospital discharge. Subgroup analyses suggest that the mortality difference between recipients of bioprosthetic and mechanical valves in the overall population is primarily driven by high-risk patients who underwent concurrent CABG. Patients undergoing isolated AVR may have little or no increased risk for early death after mechanical AVR when compared with bioprosthetic AVR. Given the low mortality rates in the isolated AVR group, we expected that the sample size of the isolated AVR cohorts would not have the statistical power to detect the difference in the death rates between the recipients of mechanical and bioprosthetic valves, if a difference existed.
These findings seem counterintuitive in that mechanical valves are recommended over bioprosthetic valves for patients who are relatively healthier and expected to live longer.3,17 Age distributions of both cohorts suggested that the medical community followed these guidelines, giving more mechanical valves to the younger than to the older patients (Table 1). If in fact no additional risk for death is associated with mechanical valves, one would expect mechanical valve recipients to have lower short-term death rates than bioprosthetic valve recipients. However, this finding is not the case for our study cohorts of elderly Medicare patients (Table 2), a finding further substantiated by our risk-adjusted analyses.
Because the most prominent mortality difference occurred on the date of surgery, the implanted mechanical valve or factors related to the mechanical valve might contribute to the difference. Bioprosthetic aortic valves are stented or stentless. Very little or no difference exists in the surgery techniques between mechanical AVR and stented bioprosthetic AVR. The stentless bioprosthetic valves accounted for only a small proportion of implanted bioprosthetic valves and require more complicated surgical techniques and longer cross-clamping and cardiopulmonary bypass times.26 Therefore, surgical techniques are unlikely to make the difference. Owing to the fundamental structural differences between mechanical and bioprosthetic valves, adverse hemodynamic interaction between the mechanical valve and a recipient’s cardiovascular system should be the direction for further investigation. The mechanical valve alone is unlikely to contribute to the excessive death.
Statistics from previous studies27-31 provided very limited information regarding early mortality after AVR surgery. Four studies27-30 investigated AVRs performed in individual institutions, all of which included fewer than 1100 patients for evaluation. None of the studies had the statistical power necessary to detect a difference in early mortality. Another early study31 was based on survey data from the 1994 Nationwide Inpatient Sample and has some fundamental differences from ours. First, their study sample was derived from 1994 national samples, and statistical inferences were obtained through survey weights; second, it included patients aged 18 to 100 years; and third, many important comorbidities were missing in multivariate analysis, such as atrial fibrillation and the use of anticoagulants or immunosuppressants.32-35
Our study has several limitations. First, as in any observational study, we could not account for clinical information not available in administrative claims databases. For example, the Medicare data do not have laboratory measures of blood pressure and left ventricular ejection fraction, which are important predictors of patient outcomes.3,35,36 In our study, only disease diagnoses were qualitatively controlled. Second, different from randomized clinical trials, observational studies often encounter confounding bias because patients were not prospectively assigned to exposures and because confounding factors were not evenly distributed between study cohorts. In an attempt to adjust for the imbalanced covariates between the 2 prosthetic valves, propensity score methods were adopted to compare the difference in early mortality between mechanical and bioprosthetic valves. However, one cannot rule out the possibility that this result was found by chance or residual confounding. Finally, our conclusion is limited by the fact that we were not able to identify the specific brand of bioprosthetic or mechanical prosthetic aortic valve. For instance, more than 9 major brands of mechanical valves exist, including caged ball, monoleaflet, or bileaflet design, with different opening angles ranging from 70° to 85°. Meanwhile, more than 17 major brands of bioprosthetic aortic valves, which are made of porcine aortic valves or bovine pericardium and may be stented or stentless, are in use.37 The ICD-9-CM codes used in this study can only differentiate between mechanical and bioprosthetic valves but cannot identify the specific type and its manufacturer. Some individual models of mechanical or bioprosthetic valves could outperform others in either category. Therefore, our study’s results should be interpreted with caution.
This nationwide cohort study in elderly Medicare beneficiaries found significant associations between the mechanical prosthetic aortic valves and the risk for death within 30 days after AVR among patients undergoing concurrent CABG. The risks for death among recipients of mechanical prosthetic vs bioprosthetic valves were highest on the date of surgery and diminished in the postoperative days. No significant association between valve type and risk for death was observed in the immediate period after hospital discharge. Further research is needed to identify the cause of this difference.
Accepted for Publication: July 5, 2014.
Corresponding Author: Dongyi (Tony) Du, MD, PhD, Office of Surveillance and Biometrics, Center for Devices and Radiological Health, Food and Drug Administration, 10903 New Hampshire Ave, Silver Spring, MD 20993 (email@example.com).
Published Online: September 15, 2014. doi:10.1001/jamainternmed.2014.4300.
Author Contributions: Drs Du and McKean had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Du, Kelman, Laschinger, Sedrakyan, Encinosa, MaCurdy, Izurieta.
Acquisition, analysis, or interpretation of data: Du, McKean, Kelman, Laschinger, Johnson, Warnock, Worrall, MaCurdy, Izurieta.
Drafting of the manuscript: Du, McKean, Warnock, Sedrakyan.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Du, McKean, Johnson, Warnock, Sedrakyan, Encinosa.
Obtained funding: Du, Kelman, Worrall, Izurieta.
Administrative, technical, or material support: Du, Laschinger, Worrall.
Study supervision: McKean, MaCurdy, Izurieta.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was performed as part of the SafeRx Project, a joint initiative of the Centers for Medicare & Medicaid Services and the US Food and Drug Administration.
Role of Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The opinions expressed in this manuscript are those of the authors and not intended to represent the opinions of the US Food and Drug Administration.
Additional Contributions: Benjamin Eloff, PhD, Center for Devices and Radiological Health, US Food and Drug Administration, helped to locate additional funding to the revision. Rahul Gondalia, MPH, Acumen, LLC, performed data analysis for the revision. No compensation was received for all persons named in the acknowledgment.