A Decade of Short-Term Outcomes in Post–Cardiac Surgery Ventricular Assist Device Implantation
Data From the Society of Thoracic Surgeons’ National Cardiac Database
Background— Previous studies showed 75% mortality before hospital discharge in patients with a ventricular assist device (VAD) placed for post–cardiac surgery shock. We examined a large national clinical database to assess trends in the incidence of post–cardiac surgery shock requiring VAD implantation, survival rates, and risk factors for mortality.
Methods and Results— We identified patients undergoing a VAD procedure after cardiac surgery at US hospitals participating in the Society of Thoracic Surgeons’ National Cardiac Database during the years 1995 to 2004. Baseline characteristics and operative outcomes were analyzed in 2.5-year increments. Logistic regression modeling was performed to provide risk-adjusted operative mortality and morbidity odds ratios. A total of 5735 patients had a VAD placed during the 10-year period (0.3% cardiac surgeries). Overall survival rate to discharge after VAD placement was 54.1%. With the earliest period (January 1995 through June 1997) used as reference, the mortality odds ratio declined to 0.72 (July 1997 through December 1999) and eventually to 0.41 (July 2002 through December 2004; P<0.0001). The combined mortality/morbidity odds ratio also declined, to 0.84 and 0.48 over identical periods (P<0.0001). Preoperative characteristics associated with increased mortality were urgency of procedure, reoperation, renal failure, myocardial infarction, aortic stenosis, female sex, race, peripheral vascular disease, New York Heart Association class IV, cardiogenic shock, left main coronary stenosis, and valve procedure (c index=0.755).
Conclusions— After adjustment for clinical characteristics of patients requiring mechanical circulatory support, rates of survival to hospital discharge have improved dramatically. Insertion of a VAD for post–cardiac surgery shock is an important therapeutic intervention that can salvage most of these patients.
Received September 21, 2006; accepted May 10, 2007.
Post-cardiac surgery shock occurs after 2% to 6% of all cardiac surgical procedures.1,2 Implantation of a ventricular assist device (VAD) is an important therapeutic option when post–cardiac surgery shock is not responsive to inotropic support and intra-aortic balloon pump insertion. The largest study to date on post–cardiac surgery VAD implantation included 965 patients and found a survival rate of <25%.3 Over the last decade, however, there have been significant advances in VAD technology and experience, and more recent studies have reported better outcomes, with up to 60% survival to discharge.4–6 Although these reports are encouraging, they are limited by small sample size, nonuniform patient selection, and influence in center experience.5,7,8
Mechanical circulatory support is an extraordinarily resource-intensive undertaking. The decision to insert a VAD in an individual patient can be difficult and is influenced by expected outcomes and patient and family wishes. This patient population has a persistently high mortality rate with conservative management. Better information on contemporary risks and predictors for those risks may better inform clinical decisions.
Editorial p 586
Clinical Perspective p 612
Therefore, using the expansive clinical experience available from the Society of Thoracic Surgeons’ National Cardiac Database (STS NCD), we sought to examine changes in patient characteristics and outcomes over the last decade in patients with post–cardiac surgery shock managed with a VAD. Our main objectives were to characterize trends in (1) the use of VADs, (2) changes in the clinical characteristics of patients who received a VAD for postoperative cardiogenic shock, (3) short-term outcomes after VAD implantation, and (4) predictors of VAD outcomes.
Clinical data for the STS NCD were collected using methods described fully elsewhere.9 Briefly, the NCD was established in 1987 as a multicenter clinical repository for quality improvement and clinical research. The STS NCD collects data from approximately two thirds of all US hospitals with cardiothoracic surgical programs and contains detailed data on patient demographics, clinical profile, and short-term outcomes for >2.6 million procedures. Data definitions are standardized, and site data coordinators receive initial and ongoing training in these definitions. Sites voluntarily submit data to the data coordinating center (Duke Clinical Research Institute) semiannually and receive site-specific feedback benchmarked against regional and national results. Accuracy of individual data elements has been validated in regional analyses, with an agreement rate >95%.9 Overall completeness of procedure reporting and mortality event reporting in patients ≥65 years of age has been validated against national Medicare claims files.10
The study population consisted of all patients in the STS NCD undergoing coronary artery bypass graft surgery and/or valve surgery between January 1995 and December 2004 who required VAD implantation after surgery. Patients who underwent cardiac transplantation during the same hospital stay were excluded. A VAD was defined as any implantable mechanical VAD (electric or pneumatic). Details on the specific VAD used and the ventricle(s) supported were not available until 2004.
The primary outcome was operative mortality defined as in-hospital death or death within 30 days of the procedure. In addition, morbidity end points examined were any reoperation, deep-sternal infection, stroke (persistent central neurological deficit >72 hours), renal failure (increase in creatinine to >2.0 mg/dL or 2-fold increase from preoperative level or initiation of dialysis), and prolonged intubation (>24 hours).
For analytical purposes, the 10-year study period was divided into four 2.5-year increments. Categorical variables are presented as percentages; continuous variables are presented as medians and interquartile ranges unless otherwise stated. For probability values, the Kruskal-Wallis test was used in Table 1 for continuous variables; the remaining variables in Table 1 and those in Table 2 used χ2 tests with an alternative hypothesis that not all rates (or distributions of rates for multilevel variables) across columns were equal. Variables with some missing values typically had low rates of missing values (not shown); however, missing values were not considered when probability values were computed. For the outcomes in Table 3, probability values are for trend across columns, determined with the Jonckheere-Terpstra (nonparametric) test for trend.
Logistic regression was used to identify multivariate preoperative risk factors for mortality and morbidity outcomes after VAD implantation. Specifically, risk-adjusted outcomes were determined from generalized estimating equation models with exchangeable correlation structure to account for correlations among clustered responses (eg, within-hospital correlations) because patients within a hospital are likely to be similar. Candidate variables for this analysis came from known predictors of bypass mortality and included >30 variables from baseline demographics, clinical factors, recent cardiovascular events, operative characteristics, type of pre-VAD cardiac procedure, and time period (see the Appendix).11 Rates of missing values were low and imputed to negative or not having the characteristic for use in logistic modeling. All analyses were performed with SAS 8.2 (SAS Institute, Cary, NC) software.
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written
Incidence of Post–Cardiac Surgery VAD Implantation
Between January 1995 and December 2004, 5735 VADS were implanted in 601 STS NCD centers. Overall, the percentage of cardiac surgical procedures requiring VAD insertion support was small, with 0.3% of all cardiac operations associated with a VAD implantation. Of centers with continuous data (n=123), the annual volume of VADs placed was <0.5% of cardiac surgical operations; 180 were placed in 1995, and 147 were placed in 2004. Most centers had implanted <50 VADs, whereas 7 sites had implanted >50 VADS over the 10-year period.
Patient and Operative Characteristics
Table 1 displays baseline characteristics in 2.5-year increments for January 1995 to June 1997, July 1997 to December 1999, January 2000 to June 2002, and July 2002 to December 2004. The median age was 63 years, with mostly white men undergoing VAD implantation. The distribution for race changed over time, with a greater proportion of blacks receiving VADs in the last time periods (P=0.009). Family history of coronary disease and median creatinine declined over time (P<0.001), whereas reporting of hypercholesterolemia increased (P<0.001). The frequency of prior coronary artery bypass graft surgery declined over time (P=0.077), but the frequency of prior percutaneous interventions increased (P<0.001) with the introduction of coronary stenting. The severity of coronary disease remained stable, with ≈25% having significant left main disease and 60% having 3-vessel coronary disease. The frequency of preoperative myocardial infarctions declined over time (P<0.001). Interestingly, most patients did not have prior heart failure before surgery, although a significant number of patients had preoperative cardiogenic shock.
Table 2 describes the operative characteristics. Most patients underwent nonelective surgery, with a significant number characterized as emergent or emergent/salvage. The frequency of aortic valve replacement was low and did not vary significantly over time. The rate of mitral valve replacement declined over time, whereas the frequency of annuloplasty increased over time. Intra-aortic balloon pump support was frequently needed preoperatively and intraoperatively, reflecting compromised hemodynamics present in the cohort early in the perioperative period. Salvage operations requiring VAD declined from 14.2% to 5.9%
The primary outcome, operative mortality, occurred in 45.9% of patients. Table 3 displays unadjusted outcomes, which show a general improvement from the 1995-to-1997 period to the 2002-to-2004 period. With the first time period of January 1995 to June 1997 used as the reference, the risk-adjusted odds of operative mortality declined by the next time period (July 1997 to December 1999) to 0.72 times the odds in the reference group, and by the final time period of July 2002 to December 2004, the odds had dropped to 0.41 times the odds of mortality of the reference group (Figure 1; P for trend <0.0001). Major complications also declined over time, except for infections and prolonged ventilation (Table 3 and Figure 2; P for trend <0.0001).
Multivariable Predictors of Mortality
Table 4 provides the independent risk factors associated with VAD operative mortality (c index=0.755). Some of the most important predictors of mortality include the urgency of the procedure, procedure characteristics, and patient characteristics such as renal dysfunction, acute myocardial infarction, and female sex. Interestingly, some characteristics such as age or diabetes mellitus were among other predictors tested that were not significant but are predictive of outcomes in areas of cardiac surgery.11 To further analyze age, we examined outcomes by age decile and showed similar outcomes with a relatively wide distribution of VAD implantations across age groups (Table 5).
In 2004, the STS NCD transitioned to a new data collection form that collected more detailed data on VAD implantation. Of the 170 VADS implanted at STS centers in 2004, 115 had detailed information with the new data collection form (Table 6). Of the 115 VADs, 32 were right sided, 51 were left sided, and 32 were biventricular devices. Most were implanted to separate from cardiopulmonary bypass as a bridge to recovery. Explantation rates were 43.3% for right VADs, 27.9% for left VADs, and 41.4% for biventricular VADs. Interestingly, discharge rates with the VAD were 28.1% for right VADs, 52.9% for left VADs, and 34.4% for biventricular VADs. Other important outcomes include embolic stroke in 6.1% of the 2004 patients: 1 right VAD patient (3.1%), 5 left VAD patients (9.8%), and 1 biventricular VAD patient (3.1%).
The present study is the largest ever to examine VAD implantation for post–cardiac surgery shock. We have observed a dramatic improvement over time in survival to discharge among patients who develop post–cardiac surgery shock. Although major complications and mortality continue to occur at high rates, this study demonstrates that continued advances in VAD technology and patient management strategies have favorably affected outcomes over the last decade. An important point is that VAD implantation has been performed across all age groups with similar outcomes, even in the very elderly.
Several possible explanations exist for the improvement in outcomes after VAD implantation. First, there have been improvements in VAD technology. Many centers have changed to VADs that are easier to implant and require less maintainance.12 The technological advances include the use of more durable pulsatile pumps instead of centrifugal pumps. Pulsatile pumps often are easier to implant because of advances in cannulation and are less likely to need to be replaced.12–14 Because of these advances, there may be fewer problems with bleeding, infections, hemolysis, and embolic events. Second, patient management also has advanced, with centers using protocols for anticoagulation and antibiotics that may have evolved over the years.12 Appropriate anticoagulation may reduce the need for reoperations for bleeding and prevent thromboembolic events in the case of subtherapeutic anticoagulation. Other medical care such as continuous hemodialysis and antibiotics and other postoperative pharmacological management such as nitric oxide may have contributed. Lastly, the advances in technology, ease of implantation, and patient care strategies may have encouraged surgeons to proceed earlier in a patient’s course for VAD implantation before multiple complications such as renal failure developed.
The present study highlights the STS NCD as a system to continuously examine national trends in cardiac surgery and use of VADs and to assess outcomes of this patient population. The largest previous study that examined outcomes of VAD implantation for post–cardiac surgery shock was published in 1992.3 This registry included 965 patients with a VAD implanted in 70 centers (44 in the United States) from 1985 to 1990 to treat post–cardiac surgery shock.3 Our study demonstrates that operative mortality has decreased dramatically since that era, with a reduction from 75% to ≈40% during the most recent time period. This improvement occurred in the face of an older population with more comorbidities than in prior studies.15 The average age was 56 years in the Pae et al3 study compared with 63 years in our study. Other risk factors were likely more prevalent compared with patients who underwent cardiac surgery during that era.15
Other more recent studies also have shown improvements but have been limited to single-institution studies with small numbers. Furthermore, important patient characteristics are not available to appropriately examine risk-adjusted trends over time. In a study of 45 patients with the ABIOMED BVS 5000 from 1994 to 2000, 31% of the patients survived to discharge.14 In a different series published in 1999, Helman et al5 reported on outcomes of 44 patients referred from 1993 to 1998 for post–cardiac surgery shock through a network of centers to Columbia-Presbyterian Medical Center. Of those 23 receiving a VAD, 26% died.
The present study underscores the importance of understanding risk factors for mortality and major morbidity after VAD implantation. The characteristics outlined in the multivariable model can be used to identify patients who are at higher risk for morbidity and mortality post–VAD implantation, allowing more informed clinical decisions. From the model, one can estimate whether patients will have a reasonable chance of survival to discharge or whether ongoing care is futile. The most important characteristics to consider were whether the index procedure was being performed as a salvage procedure or a reoperation. Other important risk factors include the presence of cardiogenic shock or an acute myocardial infarction before the index procedure. All of these factors should be considered, especially when it is not clear that a bridge to transplantation will be likely.
Age often is considered an adverse risk factor for outcomes, but our study shows that it alone should not be used to make decisions on VAD implantation in the setting of post–cardiac surgery shock. Over the last decade, VADs have been implanted across a wide-ranging age group, and outcomes have been similar across all ages in the present study. This observation likely results from patient selection because other data from the International Society for Heart and Lung Transplantation (ISHLT) Mechanical Circulatory Support Device (MCSD) Registry show that age is an important risk factor for mortality in patients primarily implanted with a VAD as a bridge to transplantation.16 Therefore, preoperative condition and comorbidities should be strongly considered in older patients before major decisions are made for treatment of post–cardiac surgery shock because this group may be overlooked.
Although operative mortality is an important end point of postcardiotomy VAD support, short-term VAD therapy may be associated with significant morbidity. From 1995 to 2004, the odds of experiencing the combined end point of mortality or major morbidity dropped dramatically, by more than a factor of 2. Previous studies have highlighted the frequent morbidity associated with VAD implantation such as neurological events, renal failure, and bleeding that also may be associated with an inability to wean from support.3 In the Pae et al3 study, renal failure had the highest odds ratio for mortality in patients successfully weaned from support. Other studies of post–cardiac surgery VAD implantation are relatively small, so it is difficult to interpret the frequency of major postoperative morbidities.7 Because of the large number of patients examined across a diversity of sites, the present study provides a more generalized view of major complications.
Despite significant improvements in post–cardiac surgery shock managed with a VAD, advances in implementation of care, VAD technology, and patient selection may further improve outcomes. Periodic examination of national trends in both VAD use and outcomes provides an opportunity to focus efforts on continuous quality improvement. As other centers have demonstrated, site-based or regional initiatives may further improve care.5 Moreover, greater implementation of protocols used in centers of excellence may improve national outcome trends. Since 2000, outcomes have remained fairly stable, with adjusted mortality odds of ≈40% of the 1995 odds (0.37 to 0.41 over the 2 most recent time periods). Although post–cardiac surgery shock may be encountered at any institution, national initiatives may be needed to organize VAD care into regions where communication links are well established between experienced VAD centers and other centers similar to trauma or acute myocardial infarction networks.
Finally, a wealth of information will soon emerge on implantable VADs through the National Heart, Lung, and Blood Institute–sponsored Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) Registry. Other data from the International Society for Heart and Lung Transplantation (ISHLT) Mechanical Circulatory Support Device (MCSD) Registry have been reported on recent trends in outcomes of VAD implantation primarily in bridge to transplantation.16 The main differences between these registries and the present study is that the STS NCD has focused on VADs as a bridge to recovery as opposed to destination (INTERMACS) and bridge to transplantation (ISHLT). In addition, the STS NCD has >120 centers reporting continuously VAD data for the last 10 years compared with the 44 centers in the last ISHLT MCSD Registry report and ≈90 Centers for Medicare and Medicaid Services–approved destination VAD centers in the INTERMACS registry. Efforts through all of these registries will be complementary and greatly facilitate our understanding of the benefits and challenges with VADs in areas of bridge to recovery, destination, and bridge to transplantation. Furthermore, future reports from the STS NCD will continue to expand the knowledge base related to VADs with a focus on the bridge-to-recovery setting by using the expanded data collection on VADs that started in 2004.
Although the present study aimed to address national trends in VAD outcomes, our data were limited to those available in the STS NCD, and exact reasons for improvement in outcomes are unknown. However, this data source represents a significant majority of US cardiac surgical centers. Our data are limited to information that is obtained clinically and reported voluntarily through a standardized case report. Over the last decade, some variables have changed slightly in definition, and others such as detailed VAD characteristics have been added only to the most recent case report form. Although likely VAD characteristics exist that are important to examine, few data sources are available to examine trends that expand over a decade. The multivariable models are limited to the preoperative and intraoperative characteristics available in the STS NCD. Future studies using the STS NCD or other registries such as INTERMACS will allow further evaluation of type of VAD and ventricles supported.
Advances in cardiac surgery, VAD implantation, and perioperative medical management have led to improvement in outcomes of post–cardiac surgery shock requiring a VAD. Risk-adjusted odds of mortality has declined to 0.41 in 2002 to 2004 from the beginning of the study period in 1995 to 1997; risk-adjusted odds of mortality/morbidity in 2002 to 2004 also is much lower, 0.48 times the odds of the 1995 to 1997 reference. Overall, VAD implantation for post–cardiac surgery shock is stable at <1% of all cardiac surgery cases despite an increase in typical operative risk factors over time. Important preoperative characteristics can be used to predict outcomes and should be considered for clinical decisions on VAD implantation, especially in cases in which futility is likely such as emergent/salvage surgery or other high-risk features. Further improvements in care of post–cardiac surgery shock can be made through national continuous quality-improvement initiatives.
Candidate variables for logistic regression models included the following: age, sex, race, smoking history, diabetes mellitus (non–insulin dependent, insulin dependent), renal failure, dialysis, hypercholesterolemia, hypertension, peripheral vascular disease, acute myocardial infarction, history of myocardial infarction, reoperative status, urgency of procedure (emergent, salvage), preoperative percutaneous coronary intervention, New York Heart Association classification, congestive heart failure, cardiogenic shock, coronary artery disease (left main disease, number of vessels), mitral insufficiency, preoperative intra-aortic balloon counterpulsation, aortic stenosis, ejection fraction, valve procedure, other procedures, and year of surgery (four 2.5-year intervals).
Sources of Funding
This analysis was funded by the Society of Thoracic Surgeons and performed at the Duke Clinical Research Institute. Dr Hernandez is a recipient of an American Heart Association Pharmaceutical Roundtable Grant (0675060N).
Dr Rogers has been a consultant to Thoratec Corp. The other authors report no conflicts.
Ferguson TB Jr, Hammill BG, Peterson ED, DeLong ER, Grover FL. A decade of change: risk profiles and outcomes for isolated coronary artery bypass grafting procedures, 1990–1999: a report from the STS National Database Committee and the Duke Clinical Research Institute: Society of Thoracic Surgeons. Ann Thorac Surg. 2002; 73: 480–489.
Post–cardiac surgery shock carries substantial risk for major mortality, and a ventricular assist device (VAD) is an option for salvaging patients with this condition. Over the last decade, substantial progress has been made with VADs, but it is unknown whether the changes in technology, implantation, or patient care have translated into improvement in short-term clinical outcomes. We examined the Society of Thoracic Surgeons’ National Cardiac Database during the years 1995 to 2004 to assess trends in the incidence of post–cardiac surgery shock requiring VAD implantation, survival rates, and risk factors for mortality. A total of 5735 patients had a VAD placed during the 10-year period (0.3% cardiac surgeries). Overall survival rate to discharge after VAD placement was 54.1%. With the earliest period (January 1995 through June 1997) used for reference, the mortality odds ratio declined to 0.72 (July 1998 through December 2000) and eventually to 0.41 (July 2002 through December 2004; P<0.0001). The combined mortality/morbidity odds ratio also declined, to 0.84 and 0.48 over identical periods (P<0.0001). After adjustment for clinical characteristics of patients requiring a VAD, rates of survival to hospital discharge have improved dramatically. Insertion of a VAD for post–cardiac surgery shock is an important therapeutic intervention that can salvage a majority of these patients.