EJCTS Click here to go to Edwards website
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to Personal Folders
Right arrow Download to citation manager
Right arrow Author home page(s):
Alexander Albert
Wael Hassanein
Ulrich Rosendahl
Ina C. Ennker
Right arrow Permission Requests
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Florath, I.
Right arrow Articles by Ennker, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Florath, I.
Right arrow Articles by Ennker, J.
Related Collections
Right arrow Valve disease

Eur J Cardiothorac Surg 2006;30:716-721
© 2006 Elsevier Science NL

Current determinants of 30-day and 3-month mortality in over 2000 aortic valve replacements: impact of routine laboratory parameters

Ines Floratha,*, Alexander Alberta, Wael Hassaneina, Bert Arnrichb, Ulrich Rosendahla, Ina C. Ennkera, Jürgen Ennkera

a Heart Institute Lahr/Baden, Germany
b Neuroinformatics Group, Faculty of Technology, Bielefeld University, Germany

Received 27 April 2006; received in revised form 11 August 2006; accepted 16 August 2006.

* Corresponding author. Address: Heart Institute Lahr/Baden, Hohbergweg 2, D-77933 Lahr, Germany. Tel.: +49 7821 725 157; fax: +49 7821 725 110. (Email: ines.florath{at}heart-lahr.com).


   Abstract
Top
 Abstract
1. Introduction
 2. Materials and methods
 3. Results
 4. Discussion
 References
 
Objective: Haematological and biochemical measurements are performed routinely before surgery to exclude organ malfunction and blood cell and coagulation abnormalities. We aimed to test routinely obtained laboratory data as factors predicting operative risk. Methods: Between 1996 and 2003, 2198 patients underwent aortic valve replacement (AVR) (908 of them with concomitant CABG) in our institute. The mean age of the study population was 69 ± 11 years (range 13–91, 43% female). Clinical and laboratory parameters based on the consolidated data mart set were evaluated by multiple logistic regression analysis. Results: The overall operative mortality (within 30 days) was 3.8% and the mortality after 3 months was 5.9%. In addition to clinical characteristics, the following laboratory values were identified as independent predictors of 30-day mortality: fasting blood glucose, antithrombine III, partial thromboplastine time and creatinine kinase. As independent predictors of 3-month mortality, the following laboratory values were indentified: fasting blood glucose, serum creatinine, antithrombine III, partial thromboplastine time, lactate dehydrogenase, sodium concentration and serum proteins. The discriminative power of the models increased if laboratory parameters were included in addition to preoperative clinical characteristics (from 0.75 to 0.79 and from 0.75 to 0.78 for 30-day and 3-month mortality, respectively). The discriminative power using the logistic EuroScore was lower (0.71 and 0.7, for 30-day and 3-month mortality, respectively). Conclusions: Laboratory parameters as objective markers for organ function and nutritional status are useful data for the prediction of 30-day and 3-month mortality after aortic valve replacement. Using modern methods of information technology, these valuable data which are stored electronically in most hospitals, can be used efficiently for research and quality control.

Key Words: Aortic valve replacement • Mortality • Risk factors


   1. Introduction
Top
 Abstract
 1. Introduction
2. Materials and methods
 3. Results
 4. Discussion
 References
 
The goals of creating predictive models for mortality after cardiac surgery are to identify modifiable risk factors and/or to provide adequate tools for risk-adjusted outcome analysis.

The majority of the models predicting the mortality risk focused on coronary artery bypass grafting (CABG) or on cardiac surgery in general [1]. In risk predicting models for cardiac surgery, patients undergoing aortic valve replacement (AVR) are underrepresented. This may underestimate their risk and hide the real risk factors in these patients. Recently, more specified studies based on large data sets have identified important risk factors and presented risk models predicting short-term mortality in patients undergoing AVR [2–7]. However, time-related outcome analysis demonstrated that 30-day mortality is an insufficient endpoint to calculate the risk of AVR, because the surgical procedure has still an impact on mortality beyond 30 days [8].

The accuracy of predictive models depends beneath the uniformity of the study group and the adequate follow-up on the quality of the underlying data. When predictive models are used as ‘common language’ by different cardiac surgical units to compare outcome, the subjective definition of variables becomes a major problem. Therefore it is advantageous to replace the subjective assessment of the patient's condition in all risk scores by objective data like laboratory values. We were interested whether routinely obtained laboratory data may improve the prediction of the perioperative risk in patients undergoing AVR.


   2. Materials and methods
Top
 Abstract
 1. Introduction
 2. Materials and methods
3. Results
 4. Discussion
 References
 
2.1 Patient population
Between March 1996 and December 2003, 2198 patients underwent AVR either alone or in combination with CABG (41%). This study population included patients receiving mechanical and bioprosthetic valve, but excluded patients with valve repair or multivalve replacement. In 109 of all patients (5%), an aneurysm of the ascending aorta was diagnosed and the ascending aorta was replaced. In Table 1 , the operative and preoperative characteristics of the patient population are shown. The routinely measured data of blood analysis is presented in Table 2 .


View this table:
[in this window]
[in a new window]
  		Table 1. Preoperative patient characteristics evaluated in logistic regression models
 

View this table:
[in this window]
[in a new window]
  		Table 2. Preoperative laboratory values evaluated in logistic regression models
 
Patients were followed either 3 or 6 months after surgery by a mailed questionnaire. Nonresponder or their relatives were contacted by telephone. The follow-up was 100% and 99% complete for 30-day and 3-month mortality, respectively.

2.2 Statistical analysis
Statistical analysis was performed using the software package SPSS (SPSS Inc., Chicago, IL, USA).

For each patient included in the present study, 38 preoperative characteristics were used from the consolidated database of our data mart system [9]. The data were based on the anaesthesiological and cardio-surgical quality assurance and the laboratory data of the clinical chemistry. All variables included into the analysis are shown in Tables 1 and 2. The missing value rate was below 5%. Missing values were completed by the mean for continuous variables and by the most frequent event for categorical variables.

By logistic regression the following categorical outcome variables were analysed: 30-day and 3-month mortality. For variable selection of both multivariate logistic regression models, the Akaike information criterion (AIC = Deviance of the model + 2 x number of included parameters) was calculated for variables showing a difference for the outcome variable with a p-value smaller or equal to 0.25. The variables were included into multiple regression models in a step wise way. The AIC was calculated each time a variable is included. The final model is reached when no more reduction in AIC is observed.

To verify the linearity assumption of logistic regression model for continuous variables, we divided the range of the continuous independent variable into groups, and for each group we plotted the logit of the outcome variable versus the group midpoint. In the case of a nonlinear increase or decrease of the logit, the continuous variable was transformed (e.g. into dichotomous) and the AIC was determined for the transformation. If the AIC of the model containing the transformed variable was lower than for the model containing the continuous variable linearly the transformed variable was included into the final model.

As recommended [10], we limited the number of predictors for each model to 10 events per variable. That means 8 predictors for 30-day mortality model (Number of deaths = 84) and 13 for the 3-month mortality model (Number of deaths = 130).

To show the improvement in the area under the ROC curve for our final model in comparison to the EuroScore model we applied the one-tailed Hanley–McNeil test [11]. The areas under the ROC curve for the presented models will be over-optimistic in comparison to the EuroScore model as it was developed in a different data set. To correct for over-optimism we generated 40 bootstrap samples using the software packages Microsoft Excel and Access. In a simulation study was shown that 40 bootstrap samples are as good as 200 (Harrell FE Jr., http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RmS/logistic.val.pdf). The areas under the ROC curves for the final models were corrected by the optimism in the apparent performance, the difference between the mean area of 40 bootstrap samples and the mean area of 40 tests in the original data.


   3. Results
Top
 Abstract
 1. Introduction
 2. Materials and methods
 3. Results
4. Discussion
 References
 
After aortic valve replacement with or without combined CABG the mortality rates were observed as 3.8% (N = 84) and 5.9% (N = 130) at 30 days and 3 months, respectively. Preoperative clinical characteristics which were identified as independent risk factors of 30-day and 3-month mortality are shown in Table 3 . The discriminative power of the models measured by the area under ROC curve (c-index) was 0.75 for both models. Including the selected laboratory values into the models increased the c-index up to 0.79 and 0.78, for 30-day and 3-month mortality, respectively. Several clinical characteristics were not further predictive and were not included in the models containing the laboratory variables as history of myocardial infarction, cardiac arrhythmias, diabetes mellitus and infection for 30-day mortality and history of myocardial infarction, infection and emergency procedure for 3-month mortality (Tables 3 and 4 ).


View this table:
[in this window]
[in a new window]
  		Table 3. Preoperative clinical characteristics identified as independent risk factors of 30-day and 3-month mortality
 

View this table:
[in this window]
[in a new window]
  		Table 4. Laboratory values additionally to preoperative clinical characteristics identified as independent risk factors of 30-day and 3-month mortality
 
The linear assumption of the logistic model was verified for all continuous variables. For several variables presented in Fig. 1a–d we suggested that transforming the continuous variable to a categorical ones may also be valid. Including the categorical variables for fasting glucose, creatininekinase and body mass index into the model for 30-day mortality instead of the continuous variables resulted in a slight decrease of the Akaike information criterion from 639 to 632. Including the categorical variables into the model for 3-month mortality, the Akaike information criterion remained relatively constant (from 847 to 845). For better medical interpretation and comparison between 30-day and 3-month mortality, we included the considered variable as categorical. All other continuous variables showed a clearly linear relation.


Figure 1
View larger version (14K):
[in this window]
[in a new window]
  		Fig. 1. Logit functions for (a) body mass index, (b) fasting glucose, (c) creatinine and (d) creatininekinase to verify the linear assumption of the logistic regression model.

 
We compared our models to models using the variables of the logistic EuroScore as suggested for prediction of the operative risk after cardiac surgery [1]. In Fig. 2 a lower c-index for the logistic EuroScore model in our study population is shown in comparison to our regression models. The areas under the ROC curves for the final models containing the laboratory variables were corrected by the optimism in the apparent performance, which was 0.02 and 0.017 for the 30-day mortality model and the 3-month mortality model, respectively. The corrected c-indices for both models (0.769 and 0.766 for 30-day mortality and 3-month mortality, respectively) were significantly larger than for the EuroScore models (p = 0.03 and p = 0.01 for 30-day mortality and 3-month mortality, respectively).


Figure 2
View larger version (11K):
[in this window]
[in a new window]
  		Fig. 2. C-index and 95% confidential interval for different models relying on clinical characteristics and laboratory values (clinical and lab model), on clinical characteristics alone (clinical model) and on the variables of the logistic EuroScore model (EuroScore model) predicting 30-day and 3-month mortality.

 

   4. Discussion
Top
 Abstract
 1. Introduction
 2. Materials and methods
 3. Results
 4. Discussion
References
 
In the present study, the predictive models for 30-day and 3-month mortality after AVR are based mainly on preoperative laboratory data.

We were able to identify preoperatively measured laboratory data as risk factors for operative mortality. Instead of renal insufficiency, preoperative myocardial ischaemia, chronic heart failure, our models included increasing creatinine concentration, the concentration of creatinine kinase, lower serum sodium concentrations. These laboratory values can be assigned to the well-known clinical risk factors used in other risk scores.

For other laboratory variables as total protein, antithrombine III (AT III) or partial thromboplastine time (PTT) the causes for the deviation from normality are miscellaneous: e.g. lower serum total protein and lower AT III levels may indicate increased risk in patients with dilution due to intravascular fluid overload [12]. But lower serum total protein can also be a marker for malnutrition, thereby increasing risk for infection [13] and mortality [14]. Nutritional support or albumin administration may be an appropriate therapy in these patients [15]. Low AT III levels lead to failure of activated clotting time to reach high levels during cardiopulmonary bypass even with adequate heparin doses with the subsequent risks. Our finding that AT III is included in the model for 3-month mortality independent from PTT and sodium indicates that lower AT III levels is an indicator for thromboembolic complications. Although no guidelines or clinical studies formally establish AT III monitoring and substitution [16], our results may support such policy.

Fasting glucose was found to be a risk factor for both the model for 30-day and 3-month mortality. As we have described recently in a series of patients receiving coronary artery bypass surgery the preoperative measurement of fasting plasma glucose (FPG) can identify patients with diabetes, according to the new definitions of the American Diabetes Association (FPG ≥ 126 mg/dl [7.0 mmol/l] = diabetes mellitus [17]), who have an increased morbidity and mortality rate during the perioperative and postoperative course [18]. Strict control of hyperglycaemia particularly in these patients is recommended.

We did not identify a low preoperative haematocrit as risk factor in contrast to the Cleveland Clinic Score. It can be a risk factor for mortality since the preoperative haematocrit correlates with the intraoperative haematocrit, which was found to be an important predictor of adverse outcome (up to 6 years) [19]. However, the correlation between preoperative to intraoperative haematocrit and subsequently the operative risk depends greatly on the perioperative management and the policy of blood products transfusion. Secondly, the haematocrit correlates with important risk factors such as female gender, low body weight and low surface area [19,20]. Thus, other important risk stratification systems such as EuroScore or Parsonnet Score [1,21], did not find that a low preoperative haematocrit but female gender is a risk factor for mortality. Analogous in the Cleveland Clinic Score female gender and low body weight were not included in the Cleveland model although they were risk factors in their univariate analysis.

To summarize, the inclusion of laboratory risk factors into predictive model yield several advantages:

1. These new laboratory risk factors are accurately defined, showing no inter-observer variability. The more complex the variable definitions (e.g. EuroScore: ‘preoperative critical state’) the greater is the inter-observer variability. Variable clinical definitions depend largely on hospital specific diagnostic standards (e.g. EuroScore: diagnosis of pulmonary hypertension by SWAN-Ganz Catheter; diagnosis of carotid stenosis by Doppler-sonography, etc.) and became inaccurate, when the screening for these diseases remains incomplete [22]. This issue may contribute to the observed variations in its accuracy at predicting risk in different surgical subgroups in two recent studies [23,24]. Thus, it seems reasonable to include only objective parameters in risk scores, where variable definitions are accurate and consistent between the units.
2. The inclusion of laboratory data improved the AIC and the discriminative power of the models. For aortic valve replacement the optimism-corrected area under the ROC curve was significantly larger for the models containing laboratory values than for the EuroScore model. With our comparably small data set (N = 2189) we were able to achieve a similar validated c-index of 0.77 for prediction of 30-day mortality after aortic valve replacement in comparison to recently published models containing much more risk factors relying on larger databases as from the Northern New England Study Group (N = 5793, c-index = 0.75) [5] and for valve replacement from the Society of Cardiothoracic Surgeons of Great Britain and Ireland (N = 32839, c-index = 0.77) [6] and from the Society of Thoracic Surgeons (N = 409904, c-index = 0.735) [7].
3. The identification of risk factors, as AT III, total proteins and fasting blood glucose, yields the opportunity to modify these parameters during the perioperative course with the intention to decrease the operative risk.


   Footnotes
 
{star} Presented at the joint 19th Annual Meeting of the European Association for Cardio-thoracic Surgery and the 13th Annual Meeting of the European Society of Thoracic Surgeons, Barcelona, Spain, September 25–28, 2005.


   References
Top
 Abstract
 1. Introduction
 2. Materials and methods
 3. Results
 4. Discussion
 References
 

  1. Nashef SAM, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R, the EuroScore study group European system for cardiac operative risk evaluation (EuroScore). Eur J Cardiothorac Surg 1999;16:9-13.[Medline]
  2. Jamieson WR, Edwards FH, Schwartz M, Bero JW, Clark RE, Grover FL. Risk stratification for cardiac valve replacement. National cardiac surgery database. Ann Thorac Surg 1999;67:943-951.[Abstract/Free Full Text]
  3. Florath I, Rosendahl U, Bauer S, Dalladaku F, Ennker IC, Ennker J. Current determinants of operative mortality in 1400 patients requiring aortic valve replacements. Ann Thorac Surg 2003;76(1):75-83.[Abstract/Free Full Text]
  4. Edwards FH, Peterson ED, Coombs LP. Prediction of operative mortality after valve replacement surgery. J Am Coll Cardiol 2001;37:885-892.[Abstract/Free Full Text]
  5. Nowicki ER, Birkmeyer NJO, Weintraub RW, Leavitt BJ, Sanders JH, Dacey LJ, Clough RA, Quinn RD, Charlesworth DC, Sisto DA, Uhlig PN, Olmstead EM, O’Connor GT. Multivariable prediction of in-hospital mortality associated with aortic and mitral valve surgery in Northern New England. Ann Thorac Surg 2004;77:1966-1977.[Abstract/Free Full Text]
  6. Ambler G, Omar RZ, Royston P, Kinsman R, Keogh BE, Taylor KM. Generic, simple risk stratification model for heart valve surgery. Circulation 2005;112(2):224-231[Epub 2005 Jul 5].[Abstract/Free Full Text]
  7. Rankin JS, Hammill BG, Ferguson TB, Glower DD, O’Brien SM, DeLong ER, Peterson ED, Edwards FH. Determinants of operative mortality in valvular heart surgery. J Thorac Cardiovasc Surg 2006;131:547-557.[Abstract/Free Full Text]
  8. Osswald BR, Blackstone EH, Tochtermann U, Thomas G, Vahl CF, Hagl S. The meaning of early mortality after CABG. Eur J Cardiothorac Surg 1999;15(4):401-407.[Abstract/Free Full Text]
  9. Arnrich B, Walter J, Albert A, Ennker J, Ritter H. Data Mart based research in heart surgery: challenges and benefit. Medinfo 2004;11:8-12.[Medline]
  10. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49(12):1373-1379.[CrossRef][Medline]
  11. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839-843.[Abstract/Free Full Text]
  12. Ng KF, Lam CC, Chan LC. In vivo effect of haemodilution with saline on coagulation: a randomized controlled trial. Br J Anaesth 2002;88(4):475-480.[Abstract/Free Full Text]
  13. Rapp-Kesek D, Stahle E, Karlsson TT. Body mass index and albumin in the preoperative evaluation of cardiac surgery patients. Clin Nutr 2004;23(6):1398-1404.[CrossRef][Medline]
  14. Bouillanne O, Morineau G, Dupont C, Coulombel I, Vincent JP, Nicolis I, Benazeth S, Cynober L, Aussel C. Geriatric Nutritional Risk Index: a new index for evaluating at-risk elderly medical patients. Am J Clin Nutr 2005;82(4):777-783.[Abstract/Free Full Text]
  15. Haynes GR, Navickis RJ, Wilkes MM. Albumin administration—what is the evidence of clinical benefit? A systematic review of randomized controlled trials. Eur J Anaesthesiol 2003;20(10):771-793.[CrossRef][Medline]
  16. Shore-Lesserson L, Gravlee GP. Anticoagulation for cardiopulmonary bypass. In: Gravlee GP, Davis RF, Kurusz M, Utley JR, editors. Cardiopulmonary bypass. Principles and practice. Philadelphia, PA, USA: Lippincott Williams &Wilkins; 2000.
  17. American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Care 2004;27:S11-S14.
  18. Lauruschkat AH, Arnrich B, Albert AA, Walter JA, Amann B, Rosendahl UP, Alexander T, Ennker J. Prevalence and risks of undiagnosed diabetes mellitus in patients undergoing coronary artery bypass grafting. Circulation 2005;112(16):2397-2402.[Abstract/Free Full Text]
  19. Habib RH, Zacharias A, Schwann TA, Riordan CJ, Durham SJ, Shah A. Adverse effects of low hematocrit during cardiopulmonary bypass in the adult: should current practice be changed?. J Thorac Cadiovasc Surg 2003;125:1438-1450.[Abstract/Free Full Text]
  20. Higgins TL, Estafanous FG, Loop FD, Beck GJ, Blum JM, Paranandi L. Stratification of morbidity and mortality outcome by preoperative risk factors in coronary artery bypass patients. A clinical severity score. J Am Med Assoc 1992;267(17):2344-2348[Erratum in: JAMA 1992;268(14):1860].[Abstract/Free Full Text]
  21. Parsonnet V, Dean D, Bernstein AD. A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 1989;79:I3-I12[Erratum in: Circulation 1990;82(3):1078].[Medline]
  22. Osswald BR, Tochtermann U, Thomas G, Vahl CF, Hagl S, HVMD Study Group Stability of the Euro-score as an identification tool for patient risk groups—dependency of the applied documentation system. Z Kardiol 2002;91(2):125-130.[Medline]
  23. Bhatti F, Grayson AD, Grotte G, Fabri BM, Au J, Jones MT, Bridgewater B. The logistic EuroSCORE in cardiac surgery: how well does it predict operative risk?. Heart 2006;17.
  24. Yap CH, Reid C, Yii M, Rowland MA, Mohajeri M, Skillington PD, Seevanayagam S, Smith JA. Validation of the EuroSCORE model in Australia. Eur J Cardiothorac Surg 2006;29:441-446.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Eur. J. Cardiothorac. Surg.Home page
A. Ioannides, A. Helmy, I. Dimitropoulos, and V. Vassiliou
What will the 'creatinine kinase' be in 2016?
Eur. J. Cardiothorac. Surg., February 1, 2007; 31(2): 333 - 333.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to Personal Folders
Right arrow Download to citation manager
Right arrow Author home page(s):
Alexander Albert
Wael Hassanein
Ulrich Rosendahl
Ina C. Ennker
Right arrow Permission Requests
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Florath, I.
Right arrow Articles by Ennker, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Florath, I.
Right arrow Articles by Ennker, J.
Related Collections
Right arrow Valve disease

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