| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Division of Cardiac Surgery, Department of Surgery, The Johns Hopkins Medical Institutions, Baltimore, Maryland
b Division of Cardiology, Department of Medicine, The Johns Hopkins Medical Institutions, Baltimore, Maryland
Accepted for publication August 6, 2009.
* Address correspondence to Dr Conte, Johns Hopkins Hospital, Blalock 618, 600 N Wolfe St, Baltimore, MD 21287 (Email: jconte{at}csurg.jhmi.jhu.edu).
Presented at the Poster Session of the Forty-fifth Annual Meeting of The Society of Thoracic Surgeons, San Francisco, CA, Jan 26–28, 2009.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterial and MethodsResultsCommentReferences |
|---|
Methods: Preoperative scores were calculated from prospective data of patients who received continuous-flow LVADs between June 2000 and May 2009. Cox proportional hazard analysis assessed the effect of preoperative variables and scores on 30-day, 90-day, and 1-year mortality. Patients were stratified by score into low- and high-risk groups. Survival was modeled using the Kaplan-Meier method.
Results: During the study period, 86 continuous-flow LVADs were implanted. The mean (± standard deviation) preoperative scores were: COL, 1.05 ± 1.59; LM, 11.9 ± 5.4; APACHE II, 15.6 ± 4.3; INTERMACS, 2.64 ± 1.01; and SHFM, 2.97 ± 1 .42. On univariate analysis, the SHFM score best differentiated low- and high-risk patients at all mortality end points; the INTERMACS and APACHE II scores were predictive for 90-day and 1-year mortality. On multivariable analysis, SHFM (hazard ratio [HR], 1.50; 95% confidence interval [CI], 1.02 to 2.21; p = 0.04) and APACHE II (HR, 1.10; 95% CI, 1.01 to 1.21; p = 0.04) predicted 1-year mortality.
Conclusions: Among the LM, COL, APACHE II, INTERMACS, and SHFM scores, the best predictor of mortality in a single institutional cohort of continuous-flow LVAD patients was the SHFM score.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentReferences |
|---|
Efforts to risk stratify preoperative LVAD patients have used available scores as well as developed novel scoring systems. The Acute Physiology and Chronic Health Evaluation II (APACHE II) and the Seattle Heart Failure Model (SHFM) scores were derived and validated in cohorts of critically ill non-LVAD patients and then applied to LVAD populations [11, 12]. Alternatively, the Columbia (COL), Leitz-Miller (LM), and Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) scores were generated from cohorts of LVAD patients to assess the risk of death after device implantation [8, 9, 13]. Studies to date have used scoring systems to risk stratify pulsatile-flow LVAD patients. However, it is unknown if these scores are applicable to patients with second-generation continuous-flow LVADs.
We believe that applying these five scoring systems to a single cohort of LVAD recipients may provide insight into which scores best predict survival. Hence, we examined the COL, LM, APACHE II, SHFM, and INTERMACS scores in our institutional cohort to investigate their capacity to predict outcomes in continuous-flow LVAD patients.
|
Material and Methods |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentReferences |
|---|
The COL Score
The COL was derived by Rao and colleagues [8] from an earlier version [14] using a cohort of 130 consecutive bridge-to-transplant VE HeartMate I LVAD patients (June 1996 to March 2001). They assigned points to five preoperative factors that predicted perioperative mortality: ventilation, 4; postcardiotomy shock, 2; previous LVAD, 2; central venous pressure exceeding 16 mm Hg, 1; and prothrombin time exceeding 16 seconds, 1. Points were assigned using relative risks. The COL was then used to stratify their population into high (> 5 points) and low (
5 points) risk of perioperative mortality (46% vs 12%, p < 0.001).
The LM Score
The LM was derived by Leitz and colleagues [9] using a cohort of 222 XVE HeartMate I LVAD patients from the Thoratec Corporation (Pleasanton, CA) Destination Therapy Registry (September 2002 to December 2005). They identified nine preoperative factors that predicted 90-day in-hospital mortality and assigned points: platelets, 148 x 103/µL or less, 7 points; albumin, 3.3 g/dL or less, 5 points; international normalized ratio exceeding 1.1, 4 points; vasodilator therapy, 4 points; mean pulmonary artery pressure of 25 mm Hg or less, 3 points; and 2 points each for aspartate aminotransferase exceeding 45 U/L, a hematocrit of 34% or less, serum urea nitrogen exceeding 51 mg/dL, and lack of intravenous inotropic support. Points were assigned using odds ratios, and patients were stratified into low (16) and high (
17) risk groups. High-risk patients had a significantly lower 1-year survival of 13% vs 69% (p < 0.001).
The APACHE II Score
The APACHE II score was developed by Knaus and colleagues [11] from an earlier version [15] using a multi-institutional cohort of 5815 critically ill patients and sought to measure the severity of disease in intensive care unit patients. It consists of 13 preoperative variables: temperature, mean arterial pressure, heart rate, respiratory rate, partial pressure of arterial oxygen or alveolar-arterial oxygen gradient if the fraction of inspired oxygen is 50% or more, arterial pH, serum sodium, serum potassium, serum creatinine, hematocrit, white blood cell count, Glasgow Coma Score, and age. Knaus and colleagues [11] found a dramatic increase in hospital deaths in congestive heart failure patients with APACHE II scores exceeding 20. Gracin and colleagues [16] later applied this cutoff to a cohort of 31 bridge-to-transplant VE HeartMateI LVAD patients and 50 patients treated with optimum medical management (September 1995 to July 1996), demonstrating that patients with APACHE II scores between 10 and 20 treated with LVADs had a mortality benefit over medical therapy patients.
The SHFM Score
The SHFM was derived by Levy and colleagues [12] from a cohort of 1125 New York Heart Association (NYHA) class IIIB or IV patients to provide estimates of survival. It uses 21 preoperative assessments, weighted by hazard ratio: age, gender, NYHA class, weight, ejection fraction, systolic blood pressure, presence of ischemic cardiomyopathy, daily furosemide equivalent dose, inotrope use, statin use, allopurinol use, angiotensin-converting enzyme use, β-blocker use, angiotensin receptor blocker use, potassium-sparing diuretic use, implantable cardioverter-defibrillator use, hemoglobin, lymphocyte percent on complete blood count differential, serum uric acid, serum cholesterol, and serum sodium [12]. Levy updated the SHFM for LVAD patients by adding two variables, intraaortic balloon pump implanted or ventilated, or both, and inotrope therapy [17].
Although risk stratifications were not defined, Levy and colleagues supplied a formula for calculation of predicted survival:
|
| (1) |
The INTERMACS Score
The INTERMACS profile was derived by Kirklin and colleagues [13] using the INTERMACS Registry, which included 511 patients from 75 institutions (314 LVAD patients). INTERMACS stratifies LVAD patients into seven levels of clinical acuity, defined as 1 ("critical cardiogenic shock"), 2 ("progressive decline despite inotropes"), 3 ("stable but inotrope dependent"), 4 ("recurrent advanced HF"), 5 ("exertion intolerant"), 6 ("exertion limited"), and 7 ("advanced NYHA III"). Multivariable analysis demonstrated that an INTERMACS score of 1 has a relative risk of 1.59 for 1-year mortality (p = 0.02).
Statistical Analysis
A univariate Cox proportional hazards analysis was performed on the five scores, their component variables, and other covariates to evaluate the risk of mortality at 30 days, 90 days, and 1 year after LVAD implantation. Censoring occurred for individuals who underwent device explantation, were lost to follow-up, or were alive at the end of the time point examined (administratively censored). A multivariable model including all five scores was constructed to evaluate the mortality end points. Scoring systems were used as continuous variables in the Cox analyses.
To further examine individual preoperative variables, a multivariable model including the eight most significant component variables on univariate analysis was constructed for each mortality end point. Given the number of observations present in our cohort, we limited our three component models to eight covariates to avoid corrupting the statistical methodology, as described by Peduzzi and colleagues [18].
For each score, a dichotomous cut point was identified allowing for stratification into high- and low-risk patients. Initial cut points were chosen based on previous studies (COL, LM, INTERMACS, and APACHE II) [8, 9, 11, 13]. We also identified cut points based on visual inspection of linear breaks in the risk of death along the continuum of the score in our cohort. For each score, a histogram was used to plot the distribution of our cohort and patients were stratified into high-risk and low-risk groups based on cut points that best fit our population. Kaplan-Meier estimates of survival comparing low-risk and high-risk patients were performed up to 1 year of device support to ensure adequate follow-up. Survival curves were compared using the log-rank test.
All means are presented with standard deviations, and hazard ratios (HR) are presented with 95% confidence intervals (CI). For all analyses, a value of p < 0.05 was considered significant. All statistical analyses were conducted using STATA 9.2SE software (StataCorp LP, College Station, TX).
|
Results |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentReferences |
|---|
|
Outcomes
Patients spent 277 ± 233 days on LVAD support, with 27 (31.4%) receiving more than 1 year of mechanical circulatory support. The 30-day, 90-day, and 1-year mortality were 10.6% (n = 9), 22.7% (n = 19), and 30.3% (n = 24), respectively.
Risk Score Distribution and Kaplan-Meier Survival
The distribution of scores and Kaplan-Meier survival by high- and low-risk groups are shown in Figure 1. If a score significantly differentiated high-risk and low-risk groups, the most statistically significant cut point is shown. Patients determined to be high-risk by INTERMACS, APACHE II, and SHFM had a significantly higher mortality rate. Meanwhile, high-risk and low-risk groups formed using COL and LM were not significantly different.
|
|
Analysis of Preoperative Variables
Univariate Cox proportional hazards analysis demonstrated preoperative variables that significantly predicted 30-day, 90-day, and 1-year mortality (Table 3). On multivariable analysis (Table 4), older age was associated with 1-year mortality, whereas a higher serum creatinine level and higher platelet count were associated with death on all three outcomes measures. None of the other variables analyzed in our three multivariable models predicted outcomes.
|
|
|
Comment |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentReferences |
|---|
Population Characteristics
The illness severity of our population is illustrated by their low ejection fractions and cardiac indices, the high prevalence of preoperative cardiogenic shock, and their nearly uniform classification as NYHA class IV. In most respects, our cohort is similar to LVAD populations in which these five scores were previously applied [8, 9, 11, 13, 17]. Our population was younger than the LM (61 years) and SHFM (67 years) populations, although it was similar in age to the COL (50 years) and APACHE II (51 years) populations. Our cohort also had a lower rate of ischemic cardiomyopathy (33%) compared with LM (65%), COL (53%), and SHFM (70%) patients. The COL population was higher acuity than patients in other studies, with 50% of patients ventilated preoperatively and 30% of patients in postcardiotomy shock, compared with 7% and 2% in our population, respectively.
Risk Score Distribution and Survival Analysis
The five preoperative scores discerned patients with an increase risk of death after implantation with varying success (Fig 1). The COL stratification (COL > 5) defined 3 patients as high risk. Survival was not significantly different in these patients. The analysis was repeated using different cut points (COL > 4, > 3, > 2, and > 1), but no significant difference was observed between high-risk and low-risk groups. Hence, the COL was a poor indicator of survival after implantation in our cohort, likely due to the extreme acuity of the COL derivation cohort.
The LM cutoff defined a greater number of patients as high-risk (n = 33). However, this definition failed to yield a significant difference in survival between low-risk and high-risk groups. Furthermore, no significant difference between groups was observed using different cut points (LM 16, 15, etc.). Hence, like the COL, the LM poorly differentiated patients with a higher risk of death after LVAD implantation in our cohort. One reason for this could be that the LM score was derived on an older population of destination therapy patients that was too dissimilar from our cohort.
Applying an INTERMACS cut point of 1, as did Holman and colleagues [19], did not yield a significant difference in survival between high-risk and low-risk groups, with a 1-year survival of 72.6% vs 50.9% (p = 0.19). This is likely due to the small number of level 1 patients in our population (12.3%). However, using the cut point of level 2 successfully discerned a high-risk subpopulation that had an absolute difference in 1-year postimplantation survival of 28.6%. INTERMACS levels 1 ("critical cardiogenic shock") and 2 ("progressive decline despite inotropes") both indicate a critical instability not present in the other profile levels. Therefore, we combined levels 1 and 2 to form our high-risk group. This appeared to be an appropriate high-risk cut point based on our cohort.
Applying an APACHE II cut point of 20 or higher, as suggested by Knaus and colleagues [11] successfully differentiated high-risk and low-risk groups in our cohort, with 1-year survival of 75.4% vs 44.6% (p = 0.02). However, a cut point of 17 or higher more dramatically discriminated between high- and low-risk subgroups, with a 1-year survival of 82.9% vs 44.4% (p < 0.001).
We defined a high-risk cut point for the SHFM as the score at which the SHFM predicted 6-month survival of less than 50%, which was 3.53. This cutoff successfully differentiated a high-risk subpopulation with an absolute decrease in 1-year survival of 37.5%. The actual 6-month survival in our SHFM high-risk subpopulation was 53.6%.
Cox Proportional Hazards Analysis of Risk Scores
On univariate analysis, the SHFM, APACHE II, and INTERMACS scores significantly predicted 1-year mortality; however, COL and LM did not predict any outcomes examined. The SHFM and APACHE II were developed on large cohorts of acutely ill patients and secondarily applied to smaller cohorts of LVAD patients. Conversely, the COL and LM were derived from small cohorts of LVAD patients. Our analysis suggests that although the APACHE II and SHFM were derived and validated on acutely ill non-LVAD patients, their large derivation cohorts allowed them to capture widely applicable predictors of death that are germane to disparate institutional cohorts of LVAD patients. Conversely, the small LM and COL cohorts yielded predictors specific to those populations, limiting their widespread applicability.
A multivariable analysis involving all scores was possible due to the lack of overlapping variables between scores. Only age, NYHA, ventilator status, serum sodium level, hematocrit/hemoglobin, prothrombin time/international normalized ratio, and preoperative inotropes were used in multiple scores. On multivariable analysis, INTERMACS did not predict outcomes despite its significance on univariate analysis, but both APACHE II and SHFM achieved significance when 1-year mortality was examined. It is difficult to examine why INTERMACS falls short in our multivariable analysis, given its lack of objective component variables. However, it is possible that the clinical picture assessed by INTERMACS overlaps with the specific preoperative variables used by the APACHE II and SHFM.
Cox Proportional Hazards Analysis of Individual Variables
Variables that significantly predicted all mortality end points on univariate analysis included serum urea nitrogen, serum creatinine, platelets, lymphocyte percentage on the complete blood cell count differential, and prothrombin time. On multivariable analysis, older age and serum creatinine level remained significant at predicting 1-year mortality. Unexpectedly, platelet count had a HR of 1.01, implying that increased levels of platelets were correlated with worse outcomes, an observation in direct opposition to LM findings. It is possible that higher levels of platelets may be associated with more thrombotic events or indicate a higher degree of inflammation at the time of implantation, leading to decreased survival.
Clinical Utility
The utility of risk scores is a function of their ability to predict mortality and their ease of use. Of the three most accurate scores, only INTERMACS can be easily assessed at the bedside. Both APACHE II and SHFM require the gathering of laboratory, physiologic, and clinical variables, making their calculation cumbersome; however, once calculated, the APACHE II and SHFM were better able to predict mortality. The ability to present survival predictions based on preoperative risk can be useful to providers in discussions with patients and families regarding decisions about LVAD implantation and appropriate postoperative expectations.
Limitations
Our study is limited by its retrospective nature and small sample size. As well, given the lack of standardized LVAD implantation criteria, our institutional cohort analysis may not be widely generalizable. Our study is also limited by the novelty of continuous-flow LVADs, and consequently, the lack of long-term follow-up for these patients. Our analysis did not stratify patients by therapeutic intent, because therapeutic intent was not a significant covariate for any of our mortality end points.
Conclusions
Our study applied five preoperative risk scores to a single institutional cohort of continuous-flow LVAD patients to determine which scores most accurately predict mortality. Although the INTERMACS and APACHE II differentiated high-risk patients and predicted 1-year mortality, the SHFM was superior in its ability to stratify continuous-flow LVAD patients into low- and high-risk groups and in its prediction of postimplantation mortality.
|
References |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  B. Worku, Y. Naka, S.-W. Pak, F. H. Cheema, O. T. Siddiqui, J. Jain, N. Uriel, R. Bhatt, P. Colombo, U. Jorde, et al. Predictors of Mortality After Short-Term Ventricular Assist Device Placement Ann. Thorac. Surg., November 1, 2011; 92(5): 1608 - 1613. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. M. Adamson, M. Stahovich, S. Chillcott, S. Baradarian, J. Chammas, B. Jaski, P. Hoagland, and W. Dembitsky Clinical Strategies and Outcomes in Advanced Heart Failure Patients Older Than 70 Years of Age Receiving the HeartMate II Left Ventricular Assist Device: A Community Hospital Experience J. Am. Coll. Cardiol., June 21, 2011; 57(25): 2487 - 2495. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. T. Owens and M. Jessup The Year in Heart Failure J. Am. Coll. Cardiol., April 12, 2011; 57(15): 1573 - 1583. [Full Text] [PDF]
|
|
|
|
|
|
J. M. Schaffer, G. J. Arnaoutakis, J. G. Allen, E. S. Weiss, N. D. Patel, S. D. Russell, A. S. Shah, and J. V. Conte Bleeding Complications and Blood Product Utilization With Left Ventricular Assist Device Implantation Ann. Thorac. Surg., March 1, 2011; 91(3): 740 - 749. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
