Article Text

PDF

Valvular heart disease
Valve type and long-term outcomes after aortic valve replacement in older patients
  1. E B Schelbert1,
  2. M S Vaughan-Sarrazin2,
  3. K F Welke3,
  4. G E Rosenthal2
  1. 1
    Divisions of Cardiovascular Diseases, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City, IA, USA
  2. 2
    Divisions of General Internal Medicine, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City, and Center for Research in the Implementation of Innovative Strategies in Practice (CRIISP), Iowa City VA Medical Center, Iowa City, IA, USA
  3. 3
    Division of Cardiothoracic Surgery, Department of Surgery, Oregon Health and Science University, Portland, OR, USA
  1. Dr Erik B Schelbert, National Heart, Lung, and Blood Institute, National Institutes of Health, 10 Center Drive, Room B1D416, MSC 1061, Bethesda, MD 20892–1061, USA; schelberteb{at}nhlbi.nih.gov

Abstract

Objective: To compare outcomes after aortic valve replacement (AVR) according to valve type specifically in older patients since valve-related risks are age-dependent; two randomised trials comparing mechanical and bioprosthetic valves found better outcomes with mechanical valves, but the samples were small and the patients were considerably younger than most who undergo AVR.

Design: Cohort study.

Setting: 1199 US hospitals.

Patients: Patients 65 years and older undergoing AVR during 1991–2003 (n = 307 054) identified through Medicare claims data.

Main outcome measures: Relative hazard ratios associated with bioprosthetic valves of (1) death (n = 131 719); (2) readmission for haemorrhage (n = 31 186), stroke (n = 25 051) or embolism (n = 5870); (3) reoperation (n = 4216); and (4) death or reoperation (reoperation free survival) in Cox regression analyses adjusting for demographic and clinical factors and hospital-level effects.

Results: Overall, 36% of AVR patients received bioprosthetic valves. Bioprosthetic valve recipients were older (77 vs 75 years, p<0.001) and generally had higher comorbidity. Bioprosthetic valve recipients had a slightly lower adjusted hazard ratios of death (HR = 0.97; 95% CI 0.95 to 0.98); readmission for haemorrhage, stroke or embolism (HR = 0.90, 95% CI 0.88 to 0.92); and death or reoperation (HR = 0.97, 95% CI 0.96 to 0.98), but a higher hazard ratio of reoperation (HR = 1.25, 95% CI 1.16 to 1.35). However, overall mortality and complication rates were more than 20 and 10 times higher, respectively, than the overall reoperation rate.

Conclusions: In older patients undergoing AVR, bioprosthetic valve recipients had slightly lower risks of death and complications, but a higher risk of reoperation. Given the low reoperation rate, these data suggest that bioprosthetic valves may be preferred in older patients.

View Full Text

Statistics from Altmetric.com

Aortic valve replacement (AVR) is the most common surgery for valvular heart disease.1 2 Yet, empirical data to guide selection of the appropriate type of prosthetic valve are limited. This is especially true in older patients, for whom wide treatment variation has been reported.3 4 Only two small randomised controlled trials involving a total of 605 patients have compared long-term outcomes in patients receiving bioprosthetic or mechanical valves. The larger Veterans Affairs Cooperative Study on Valvular Heart Disease trial showed better overall survival in patients receiving mechanical valves,5 whereas the smaller Edinburgh Heart Valve Trial showed better reoperation-free survival, but no differences in overall survival.6 Both trials demonstrated superior durability with mechanical valves compared to bioprosthetic valves, but at the expense of higher incidence of bleeding, given the need for perpetual anticoagulation.

While generalisability of trial data is often presumed,7 a number of methodological issues raise potential concerns about the applicability of these two earlier valve replacement trials to contemporary practice. First, a number of advances may have occurred in heart valve prostheses design and operative techniques to improve outcomes over 25 years since these trials were initiated.8 Second, most of the study population in these trials was under 60 years of age9 10 —in contrast to the majority of US patients receiving AVR, of whom only 28% are under the age of 60.11 Although the trials demonstrated that valve-related risks varied with age,12 13 the pertinence of these trials to older patients has been debated.8 1417 Third, treatment3 4 18 and outcome19 can vary substantially across individual providers and hospitals, and it is uncertain if surgical outcomes in the hospitals in which these trials were conducted reflect usual practice.

To address the limitations of the earlier randomised trials, we conducted a large scale, long-term observational study to compare outcomes of bioprosthetic AVR and mechanical AVR in a representative national sample of older patients. We specifically compared (1) survival, (2) reoperation-free survival and (3) readmission for bleeding, stroke or embolism according to valve type in over 300 000 Medicare beneficiaries undergoing AVR from 1991–2003.

METHODS

Data sources

The study utilised Medicare Provider and Analysis Review (MedPAR) Part A public use data files, which contain data from the UB-92 hospital discharge abstracts for all Medicare Part A beneficiaries who are discharged from acute care hospitals. These data included demographics; patients’ zip code and state of residence; primary and secondary diagnoses and procedures, as captured by International Classification of Diseases, 9th Clinical Modification (ICD-9-CM) codes; admission and discharge dates; disposition at the time of hospital discharge; admission source (for example, transfer from another hospital, emergency department); surgical priority (urgent and emergent admission); and a six-digit hospital identifier. MedPAR records were matched to Medicare enrolment data to obtain dates after hospital discharge (see appendix A). Zip-code level socioeconomic measures (for example, per-capita income), were determined for each patient by linking zip code of residence with data from the 1990 and 2000 US census.20

Patient sample

Patients undergoing AVR between 1 January 1991 and 30 June 2003 who were 65 years and older (n = 328 841) were identified on the basis of specific ICD-9-CM procedure codes (35.21 and 35.22). Patients simultaneously receiving a prosthetic mitral valve (n = 21 541), or pulmonic valve (n = 65) or tricuspid valve (n = 181) were excluded from analysis. These exclusions left a final study cohort of 307 054 patients who underwent AVR in 1199 US hospitals.

Data elements

In accordance with the previous randomised trials,5 6 we examined overall survival, reoperation and reoperation-free survival (that is, survival with the original prosthesis intact). Survival time was determined in days using the date of death available from Medicare enrolment files. Time to the first reoperation (in days) was determined by identifying hospitalisations for repeat AVR subsequent to the index AVR for each patient. Additional outcomes included readmission for a valve related complication, defined as bleeding, systemic embolism or stroke. Appendix B provides a list of ICD-9-CM codes used to identify each of the complications. To examine the robustness of findings related to complications, we also considered alternative ICD-9-CM code definitions for each complication.

Data on mortality and readmissions were available for all Medicare beneficiaries to 31 December 2003, the end of the observation period. Follow-up times ranged from 184 days (for people undergoing AVR on 30 June 2003) to 4747 days (for people undergoing AVR on 1 January 1991).

Statistical analyses and risk adjustment

Demographic characteristics, socioeconomic status and prevalence of coexisting conditions were compared for patients who received bioprosthesis and mechanical valves using the χ2 statistic (for categorical variables) or Wilcoxon rank-sign test (for continuous variables). Comparisons of crude numbers of events between groups ignore variation in follow-up time and by themselves are not meaningful. Thus, incidence rates for each outcome were standardised to account for the variable duration of follow-up, expressed as events per person-year and defined as the total number of events divided the by the sum of the duration of follow-up periods for all subjects in the sample. Differences in times to events were compared using the log-rank test.

Multivariable survival analysis was conducted using Cox regression models that included patient characteristics representing risk factors for adverse outcomes and right-censored for death. Candidate variables for the models included patient demographics, admission source, zip-code level socioeconomic status, surgical priority (emergent, urgent, elective) and coexisting clinical conditions as defined by Elixhauser et al.21 Stepwise selection identified variables that were independently associated with each end point (criteria for entering the model and remaining in the model after further addition/removal of other variables: p<0.01).

Age was expressed as a continuous variable in the regression models. Race was included in models using indicator variables to represent black and other non-white race, with white race as the referent category. Patient socioeconomic status was measured by the median household income and median house value of the patient’s residence zip code. Surgical priority was expressed using indicator variables for urgent and emergent admissions, relative to elective admissions. Admission source was expressed as indicator variables for patients transferred to the hospital from another acute care facility and for patients admitted through the emergency department, with a referent category that primarily included patients referred by a physician. A small proportion of patients could not be matched to zip-code level socioeconomic data (n = 7958, 2.5%). Therefore, models were run both excluding patients with missing socioeconomic data and including these patients, but replacing their missing zip-code level variables with the mean values for the sample in the risk-adjustment models. Since results were nearly identical, only the latter are reported.

Variables included in the models are shown in the Results section. An indicator variable for the use of a bioprosthetic valve was added to the final risk-adjustment model for each outcome. ICD-9-CM procedural codes were used to classify patients as either undergoing bioprosthetic valve AVR (35.21) or mechanical valve AVR (35.22). To investigate the possibility that the hazard ratios associated with the use of bioprosthetic versus mechanical valves varied by age, we conducted additional analyses for subsets of patients stratified by age category: 65–69 years (n = 58 815; 19%), 70–74 years (n = 81 343; 26%), 75–79 years (n = 87 210; 28%), 80–84 years (n = 57 535; 19%) and 85 years and above (n = 22 151; 7%).

In order to guide patients and their physicians at time of surgery when a binary choice of one valve type or another must be made, these analyses sought to summarise the overall adjusted hazard ratio22 of one valve type relative to the other using Cox regression models. In general, the hazards of one valve type relative to another are known to be non-proportional and vary as a function of time after valve implantation. For example, bioprosthetic valve recipients face increasing risks of structural deterioration relative to mechanical valve recipients,1 2 and since bioprosthetic valves are typically reserved for patients with limited life expectancy,1 2 bioprosthetic valve recipients may also have a higher short-term mortality ratio earlier after valve implantation. However, we followed the precedent set by the original valve trials that summarised the overall risks of one valve type to another,5 6 and thus ignored potential interactions of valve type with time (that is, the proportional hazard assumption). We report the “averaged” hazard ratios22 of bioprosthetic valves relative to mechanical valves for the 13 year time period following AVR.

All Cox regression models adjusted for clustering of patients within hospitals by stratification at the hospital level through the SAS procedure PHREG.22 This approach estimates model covariates conditional on hospital, and thus accounts for variation at the hospital level (for example, hospital volume) that might influence outcomes.22 All analyses were performed using SAS for Windows Version 9 (SAS Institute). Approval to conduct the study was obtained from the University of Iowa institutional review board.

RESULTS

Overall, 36% of study patients received bioprosthetic valves, although the use of bioprosthetic valves increased over time, rising from 18% in 1991 to 59% in 2003 (summarised in table 1). On average, patients receiving bioprosthetic valves were two years older than patients receiving mechanical valves, and were somewhat more likely to be white and to reside in areas with higher median household income. Patients receiving bioprosthetic valves were also more likely to undergo concurrent coronary bypass surgery and have a number of comorbid conditions, including previous myocardial infarction, chronic heart failure, chronic renal failure, cerebrovascular disease, peripheral vascular disease, diabetes, chronic obstructive lung disease, history of metastatic and non-metastatic cancer, peptic ulcer disease, coagulopathy, depression and iron deficiency anaemia.

Table 1 Baseline characteristics in bioprosthetic and mechanical valve recipients

Unadjusted incidence rates for death and reoperation (as expressed by events per 100 person-years of follow-up) were modestly higher in bioprosthetic valve recipients (table 2). However, incidence rates for readmission for haemorrhage, stroke and embolism were lower in bioprosthetic valve recipients. Rates for death were roughly two times higher than rates for readmission for any complication and more than 20-fold greater than rates of reoperation. The low rates of reoperation relative to rates of death for bioprosthetic valve recipients are shown in figure 1.

Figure 1 Kaplan-Meier curves showing the unadjusted rates of death and rates of reoperation for recipients of bioprosthetic valves.
Table 2 Numbers of events and incidence rates (events per 100 person-years of follow-up) in bioprosthetic and mechanical valve recipients

In analyses limited to patients undergoing AVR during 1991–3 (n = 60 357), patients for whom a minimum of 10 years of potential follow-up were available, bioprosthetic valve recipients had higher overall mortality than mechanical valve recipients (74% vs 70%; p<0.001) and higher rates of reoperation (3.1% vs 2.3%; p<0.001). For both groups rates of death were more than 20 times greater than rates of reoperation.

In Cox regression analyses, the unadjusted overall hazard ratios of death and death or reoperation were 7% higher for patients receiving bioprosthetic valves, relative to patients receiving mechanical valves. However, after adjusting for patient risk factors and accounting for hospital-level variation, the hazard ratios of death and death or reoperation were 3% lower for bioprosthetic valve recipients, although the adjusted hazard ratio of reoperation alone was 25% higher for bioprosthetic valve recipients. These results were nearly identical in analyses that were limited to patients receiving concurrent coronary artery bypass surgery. For example, the adjusted hazard ratio of death were 4% lower in bioprosthetic valve recipients (HR = 0.96; 95% CI, 0.94 to 0.98).

Unadjusted and adjusted hazard ratios of readmission for haemorrhage, stroke and embolism were also lower for bioprosthetic valve recipients (table 3); Kaplan-Meier curves adjusted for the covariates in the Cox regression models show outcomes according to valve type graphically and are displayed in figures 2 and 3. For example, adjusted hazard ratios ranged from 7% lower for stroke to 17% lower for haemorrhage. Additional analyses using alternative definitions for each complication yielded nearly identical findings (results not shown).

Figure 2 Kaplan-Meier curves showing (A) the adjusted rates of death according to valve type, and (B) adjusted rates of death or reoperation according to valve type.
Figure 3 Kaplan-Meier curves showing the adjusted rates according to valve type of: (A) any complication (excluding death or reoperation); (B) haemorrhage; (C) stroke; (D) embolism.
Table 3 Unadjusted and adjusted* hazard ratios of death and complications in bioprosthetic valve recipients relative mechanical valve recipients

The modest survival and reoperation-free survival benefits for bioprosthetic valve recipients in the entire cohort were not observed in patients aged less than 70 years (table 4). In patients 65–69 years, the adjusted hazard ratios of death, reoperation and death or reoperation were 4%, 31%, and 6% higher, respectively, for bioprosthetic valve recipients, relative to mechanical valve recipients. However, the risks of readmission for haemorrhage, stroke and embolism were 28%, 12% and 79% lower, respectively, for bioprosthetic valve recipients. In analyses of other age groups (70–74, 75–79, 80–84 and 85 years and older), adjusted hazard ratios of death were modestly lower for bioprosthetic valve recipients for all groups, although no clear trend in the magnitude of the hazard ratio was observed across age ranges. The hazard ratio of reoperation was higher only for patients aged 70–74 years, and differences between bioprosthetic and mechanical valve recipients in the hazard ratios of readmission for haemorrhage, stroke and embolism declined as age increased.

Table 4 Adjusted hazard ratios (95% CI) of death and complications in bioprosthetic valve recipients relative to mechanical valve recipients

To determine if the modestly higher mortality in bioprosthetic valve recipients aged less than 70 years was related to the risks associated with reoperation, we conducted additional analyses excluding patients who died within 30 days of reoperation. Such deaths represented only 1% of all deaths in patients aged 65–69 (217 of 20 823). Excluding these patients, the hazard ratio of death in bioprosthetic valve patients remained 5% higher (HR = 1.05, 95% CI 1.01 to 1.09, p = 0.02); thus, the higher mortality associated with bioprosthetic valve use in patients aged 65–69 was not attributable to higher 30-day mortality following repeat aortic valve operation.

DISCUSSION

The current study represents the first large-scale analysis comparing outcomes of patients receiving bioprosthetic or mechanical valves for AVR. Using a national database of Medicare beneficiaries undergoing AVR from 1991 to 2003, the study found that older bioprosthetic valve recipients had a slightly lower risk of death and a lower risk of hospital readmission for three potential complications of AVR—haemorrhage, stroke and embolism. Although bioprosthetic valve recipients had a higher relative risk of reoperation, the overall absolute risk of reoperation was small compared to the absolute risk of death. While recipients of bioprosthetic valves were older and had greater comorbidity, this study demonstrates for the first time, that after accounting for such factors, bioprosthetic valves may have modest benefits in older patients undergoing AVR, particularly those patients aged 70 years and older.

It is notable that death occurred far more frequently than reoperation. Among patients for whom 10 years or more of potential follow-up were available, more than 70% died during the follow-up period and less than one in 30 underwent a reoperation. The risk of reoperation is considered the main disadvantage to using bioprosthetic valves,1 2 and current guidelines provide a benchmark reoperation rate of 10% or less after 10 years for patients over age 65. The incidence of reoperation for bioprosthetic valves in this study was well below the guideline threshold. In fact, comparisons of reoperation rates after 11–13 years for recipients of bioprosthetic or mechanical valves (3.1% vs 2.3%) differed significantly from those reported by the VA Valve trial (9% vs 0%, respectively)5 at 15 years for patients over age 65. These data suggest that older mechanical valve recipients may also require reoperation (for example, from a perivalvular leak), and that the rates of reoperation in mechanical valve recipients are higher than those suggested by trials conducted in younger patients in a relatively small number of medical centres.

In subgroup analyses, the results observed for patients less than 70 years were somewhat mixed. While the use of bioprosthetic valves in such patients was associated with substantially lower risks of readmission for AVR complications, particularly with regard to haemorrhage, they also had slightly higher hazard ratios of death, which could not be directly attributed to reoperation-related complications. Factors underlying the slight increase in risk of mortality associated with bioprosthetic valves in patients older and younger than 70 are uncertain.

Yet, in interpreting the current findings, it is important to consider the potential limitations of claims data in which the prevalence of important diagnostic and prognostic variables may be underestimated when compared to clinical databases.23 Moreover, important prognostic factors such as medications, left ventricular ejection fraction, electrocardiographic findings, laboratory test results and health and functional status cannot be obtained from claims data. Yet, given the recommendation that bioprosthetic valves be reserved for patients with the shortest life expectancy,1 2 10 24 and the inherent limitations in risk adjustment using claims data, it is probable that unmeasured severity would be greater in bioprosthetic valve recipients. Therefore, analyses based on risk-adjustment models derived from clinical data may have found larger survival differences that further favour bioprosthetic valves, and our analyses may actually underestimate the “true” benefit associated with bioprosthetic valve use. The increased risk of death in patients under age 70 was not related to 30-day mortality following repeat valve operation (that is, valve redo surgery). Therefore, we suspect that unmeasured patient characteristics related to overall frailty in those selected to receive bioprosthetic valves underlie the increased mortality risk observed in these patients. We do not believe our data conflict with ACC/AHA Guidelines that generally propose an age cut-off of 65 years.1 2

The findings of slightly higher overall survival and reoperation-free survival overall in older patients with bioprosthetic valves contrast with the findings from earlier randomised trials.5 6 The differences may reflect the generally younger populations studied in the previous trials and may indicate that older patients have different risk-benefit ratios than younger patients.1 2 For example, older patients may have higher side effects from the long-term anticoagulation13 required to mitigate the thromboembolic risks posed by mechanical valves. Given their lower incidence of reoperation, older patients also have less exposure to the additional surgical risks from reoperation. In addition, bioprosthetic valves may be subject to less structural deterioration in older patients than in younger patients.1 2 Longevity of more modern bioprosthetic valves may also be greater than the older designs used in the earlier trials.8 25 Thus, our empirical data on the benefits of bioprosthetic valves are in agreement with results derived from simulation models,26 27 observational data28 and guidelines.1 2 Lastly, the current findings may suggest that the results of previous trial data, which were conducted in a select group of medical centres, may not be generalisable to broader populations of patients undergoing AVR in all US hospitals.

Our study has several important limitations. First, it was a retrospective observational study that relied on administrative data and may be subject to selection bias and confounding by unmeasured severity of illness that may be correlated with the use of different valves. For example, we were unable to identify patients with preoperative atrial fibrillation in whom receipt of a mechanical valve would be more likely because many would require persistent anticoagulation regardless of valve type. However, as noted previously, our findings are likely to underestimate the potential survival benefits of bioprosthetic valves, given that clinicians are apt to use bioprosthetic valves in patients with limited anticipated lifespans.10 24 Second, we could not reliably ascertain other important end points, such as valve deterioration, cardiovascular symptoms, functional status or decrements in quality of life associated with the use of anticoagulation therapy for mechanical valves and the required monitoring of anticoagulant dosages. Finally, the reliability of the coding for bioprosthetic and mechanical valves in administrative data has not been definitively established. Our findings, however, agree with previous reports that cite frequent use of mechanical valves in older patients.26 29 30 Moreover, if misclassification of valve type did occur, the pattern would probably be random and serve to minimise the magnitude of differences between bioprosthetic and mechanical valves.

Despite these limitations, the current large-scale, national analysis provides important information to guide valve-type selection for older patients. Published guidelines1 2 generally recommend bioprosthetic valves for older patients. Although our data indicated the use of bioprosthetic valves increased between 1991 and 2003, substantial variation in prosthetic heart valve selection for older patients still remains,3 4 indicating a lack of consensus among clinicians. The current findings underscore the potential benefits and the lack of harm in using bioprosthetic valves in older patients—especially those over age 70—and generally support the selection of bioprosthetic valves as a marker of quality in older patients undergoing AVR.

AppendixAppendix A

Although not originally intended for research, administrative databases have played a valuable part in health services research.1 2 Administrative databases clearly offer important advantages. They are relatively inexpensive and readily available; they generally cover large groups of people, they are often available for a number of years, and one can measure some aspects of the clinical status of patients.

The platform for nearly all administrative databases of hospitalised patients is the Uniform Hospital Discharge Data Set (UHDDS), originally formulated in 1972.3 The UHDDS provides a uniform but minimum dataset that facilitates investigation of cost and quality of short-term hospital services across populations at local and national levels. The most recent format is the 1992 Uniform Bill (UB-92), which is now used for 95% of hospital claims.4 The UB-92 includes patient identifiers, descriptors of hospitalisation, charge and billing data, and space for nine diagnoses codes and six procedure codes (generally recorded using the International Classification of Disease Coding Manual, version 9 (ICD-9-CM), and procedure dates.

Medicare Provider Analysis and Review (MedPAR) Part A files contain data from claims submitted by hospitals, hospices and skilled nursing facilities. Each record in the Part A files represents an inpatient stay and contains data from the UB-92 discharge abstract for the admission. Data are initially transmitted by facilities to local fiscal intermediaries, which forward the data to one of nine national processing centres. The processing centres: (1) check the beneficiaries’ eligibility status; (2) edit the claim for consistency, remaining beneficiary benefits and deductibles; and (3) authorise the fiscal intermediary to pay or reject the claim or to gather additional documentation. Once the claims are paid or rejected. The processing centres then submit the claims to the Center for Medicare and Medicaid Services (CMS), where the data are entered weekly into the National Claims History 100% Nearline File, from which the MedPAR files are derived.

Administrative data do have important drawbacks, most notably in their use of ICD-9-CM codes to measure clinical status. For example, ICD-9-CM codes may not capture laboratory values or physical examination findings that have prognostic value or may not differentiate pre-existing clinical conditions from conditions that developed during hospitalisation and, thus, represent complications of care. There are also concerns about errors in ICD-9-CM coding using information in the patient’s medical records.5 6 Studies that have formally investigated the accuracy of ICD-9-CM codes, relative to information in patients’ medical records711 have found varying levels of agreement between clinical data sources and administrative data. While agreement is generally excellent for major procedures (for example, CABG, PCI), agreement may be poorer for some comorbid conditions, leading some to recommend caution when using administrative data to make inferences about quality of care.

Appendix B

ICD-9-CM code to define end points for complications

Readmission for haemorrhage

Spontaneous intracranial haemorrhage (430, 431, 4320, 4321, 4329), intracranial haemorrhage following head trauma without skull fracture (85200–3, 85205–6, 85209, 85220, 85221–9, 85240, 85241–2, 85245–6), haemorrhage along the GI tract (45620, 4590, 5307, 53082, 53100–1, 53120, 53140, 53141, 53160–1, 53200–1, 53220–1, 53240–1, 53260–1, 53300, 53320, 53340, 53360, 53400, 53440, 53441, 53460, 53783, 53784, 56202, 56203, 56212, 56213, 5693, 5780, 5781, 5789), haematuria* (5997*), and acute post-haemorrhagic anaemia (2851).

Readmission for stroke

Spontaneous intracranial haemorrhage (430, 431, 4320, 4321, 4329), embolic stroke (43411), cerebral thrombosis (43401), unspecified cerebral artery occlusion (43491), transient ischemic attacks* (4350*, 4351*, 4353*, 4358*, 4359*), occlusion and stenosis of the precerebral arteries from embolism, narrowing, obstruction, thrombosis (43301, 43311, 43321, 43331, 43381, 43391).

Readmission for embolism

Cerebral embolism (43411), systemic embolism (44421, 44422, 44481, 44489, 4449, 5570, 59381), retinal artery occlusion* (36331–2*), transient ischemic attacks (4350, 4351, 4353, 4358, 4359), occlusion and stenosis of the precerebral arteries from embolism, narrowing, obstruction, thrombosis (43301, 43311, 43321, 43331, 43381, 43391).

*Codes marked with an asterisk were not included in the complication rate using “Modified Definitions”.

REFERENCES

View Abstract

Footnotes

  • EBS analysed the data and takes responsibility for its accuracy and integrity.

  • Funding: GER and MSV-S are supported by a grant (HFP 04–149) from the Health Services Research and Development Service, Veterans Health Administration, Department of Veterans Affairs. EBS was supported by a Cardiovascular Interdisciplinary Fellowship (HL 07121) from the University of Iowa Division of Cardiovascular Diseases and the Cardiovascular Research Center, where he was an Iowa Scholar in Clinical Investigation Program K30 trainee (K30HL04117–01A1).

  • Competing interests: None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.