J Thorac Cardiovasc Surg 2003;126:2044-2051
© 2003 The American Association for Thoracic Surgery
Surgery for acquired cardiovascular disease |
Gender profiling in coronary artery bypass grafting
Colleen Gorman Koch, MD, MSa,*,
Farah Khandwala, MSb,
Nancy Nussmeier, MDc,
Eugene H. Blackstone, MDd
a Department of Cardiothoracic Anesthesia (G-3), Cleveland, Ohio, USA
b Department of Biostatistics, The Cleveland Clinic Foundation, Cleveland, Ohio, USA
c Department of Anesthesia, Texas Heart Institute, Houston, Tex, USA
d Department of Thoracic and Cardiac Surgery, The Cleveland Clinic Foundation, Cleveland, Ohio, USA
Received for publication April 16, 2003; revisions received June 4, 2003; accepted for publication June 13, 2003.
* Address for reprints: Colleen Gorman Koch, MD, MS, Department of Cardiothoracic Anesthesia (G-3), The Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH 44195, USA
kochc{at}ccf.org
 |
Abstract
|
|---|
OBJECTIVE: Although women are reported to be at increased risk of poor outcome after coronary artery bypass grafting, female gender may simply be a marker of a high-risk profile. Therefore, the objective of this study was to characterize the difference between the female and male profiles of patients presenting for coronary artery bypass grafting.
METHODS: From January 1993 to June 2002, 15,597 patients underwent isolated coronary artery bypass grafting at a single institution. Multivariable logistic regression was used to develop a model of female gender.
RESULTS: Of 15,597 patients, 3596 (23%) were women. Eighteen variables were predictive of the female gender profile, including shorter stature, increased weight, more hypertension, insulin-treated diabetes mellitus, heart failure, and higher triglyceride and high-density lipoprotein cholesterol levels. Hematocrit, bilirubin, and creatinine values were lower in women compared with men.
CONCLUSIONS: The preoperative profiles of women and men undergoing coronary artery bypass grafting are dissimilar. Statistical modeling techniques provide a unique perspective on the preoperative profile of the female patient, who is known to be at a higher risk undergoing coronary artery bypass grafting.
When examining gender differences in outcome after cardiac surgical procedures, women have been typically profiled as high-risk primarily because of a disproportionate distribution of preoperative risk factors associated with adverse outcomes. There are numerous studies that tabulate differences in the gender distribution of preoperative baseline variables.1-12 This study characterizes the difference between the female and male profiles of patients presenting for coronary artery bypass grafting (CABG).
|
Methods
|
|---|
Patient population and data collection
From January 1993 to June 2002, 15,597 patients underwent isolated CABG at the Cleveland Clinic Foundation; patients undergoing concomitant valvular procedures were excluded. Data were prospectively collected for 64 preoperative variables, including baseline demographics, clinical history, laboratory values, and cardiac catheterization findings, and entered into institutional computer databases by individuals trained in database collection methods. Institutional review board approval was obtained from the Cleveland Clinic Foundation to perform research analyses on these institutional databases.
Statistical methods
The analysis identified a parsimonious set of characteristics that best differentiated men from women undergoing CABG, thus creating a gender profile.
Preliminary analysis
Initial exploratory analyses were performed by calculating summary statistics for each variable. The only variables to have a considerable amount of missing data were triglyceride, high-density lipoprotein (HDL), and low-density lipoprotein (LDL) cholesterol levels, with approximately 44% missing in each. These variables were examined further, but no obvious missing patterns were found. Thus, noninformative imputation was used, and missing value indicators were included in the final model. Univariable statistics were also initially computed to compare men and women using the Student t test for continuous variables and the chi-square test or Fisher's exact test statistics for categoric variables.
Before modeling the data, continuous and ordinal variables were plotted against the logit of being female to identify possible transformations necessary to linearize the relationship.
Multivariable analysis
Bootstrap aggregation (bagging) was used to identify variables and their candidate transformations.13 Coronary branches (mid, proximal, and distal) were excluded from the bagging analysis (discussed next). A total of 200 data sets of 15,597 were randomly selected with replacement and automatically analyzed by stepwise regression with an entry significance level of 0.10 and a significance level of 0.05 to stay in the model. Variables chosen in at least 50% of the models were considered for the final parsimonious model.
Aggregation step
To account for correlation among covariates, a cluster analysis was performed by investigating variables that may instinctively cluster together, such as height, weight, body surface area, and body mass index. Exploratory analysis was performed within clusters by plots and intermediate modeling to identify any interactions and distinguish which parameters and transformations have the most distinct relationship with the probability of being female.
Finally, a parsimonious model was determined with the selected covariates and missing value indicators for the variables where applicable. The model was further reduced by backward elimination of P values greater than .05. It is worth noting that although albumin appeared in 85% of the bootstrap models, it did not remain in the final model. Last, the coronary branches were successively added, yet none were found to significantly contribute to the quality of the model. All results were computed using SAS 8.2 software (SAS Institute Inc, Cary, NC).
|
Results
|
|---|
Baseline characteristics
A summary of the distribution of baseline characteristics according to gender is presented in Tables 1 and 2. Of the 15,597 patients, 3596 (23%) were women and 12,001 (77%) were men. Women were older (Figure 1) and shorter (Figure 2), weighed less (Figure 3), and had a smaller body surface area and body mass index compared with men. The occurrence of many comorbid conditions was more frequent in women, such as more hypertension and insulin-treated diabetes, as well as a history of heart failure, prior stroke, renal disease, and peripheral vascular disease. Women also had lower preoperative hematocrit levels, lower calculated red blood cell mass (Figure 4), less favorable average LDL cholesterol and triglyceride profiles, and higher HDL cholesterol levels (Figure 5). Preoperative serum creatinine measurements were lower in women (Figure 6). Compared with men, women had a more unstable presentation reflected by a higher prevalence of unstable angina, preoperative intra-aortic balloon pump usage, and emergent surgery. Women more frequently received beta blockers and angiotensin-converting enzyme inhibitors. Men had a higher prevalence of smoking, alcohol use, and abnormal left ventricular ejection fractions, as well as more extensive coronary artery disease. Men also presented more frequently for coronary reoperations.
View larger version (8K):
[in this window]
[in a new window]
|
Figure 1. The gender distribution for age in patients undergoing CABG. Women are, on average, older than men. The numbers above the data points on the plot represent the sample size per bin.
|
|
View larger version (8K):
[in this window]
[in a new window]
|
Figure 2. The gender distribution for height. Women are of significantly shorter stature compared with men. The numbers above the data points on the plot represent the sample size per bin.
|
|
View larger version (7K):
[in this window]
[in a new window]
|
Figure 3. The distribution of weight demonstrates that women characteristically weigh less than men. The numbers above the data points on the plot represent the sample size per bin.
|
|
View larger version (8K):
[in this window]
[in a new window]
|
Figure 4. A large percentage of women have lower estimated red blood cell mass. The numbers above the data points on the plot represent the sample size per bin.
|
|
View larger version (7K):
[in this window]
[in a new window]
|
Figure 5. In this study, women had significantly higher HDL levels compared with men. The numbers above the data points on the plot represent the sample size per bin.
|
|
View larger version (8K):
[in this window]
[in a new window]
|
Figure 6. Overall serum creatinine values were lower in women compared with men undergoing CABG. The numbers above the data points on the plot represent the sample size per bin.
|
|
Parsimonious model
The parsimonious analysis identified 18 variables that reflected the female profile (Table 3). This parsimonious model had high discriminatory ability to distinguish females from males, yielding a c statistic of 0.975. Shorter height, increased weight, more hypertension, insulin-treated diabetes, and history of heart failure were significantly associated with the probability of female gender. The profile of women was further characterized by increased HDL levels, increased triglycerides, lower serum bilirubin, and a reduced red blood cell mass. In contrast, the male profile was characterized by more smoking, alcohol consumption, reduced left ventricular ejection fractions, and more extensive left main and circumflex coronary artery disease.
|
Discussion
|
|---|
Gaining insight into poor outcomes after coronary artery revascularization procedures requires close scrutiny of the preoperative risk profile in those patients considered highest risk for the procedures. Female gender has been associated with increased in-hospital morbidity and mortality after CABG primarily because of the unfavorable preoperative risk profile in women. Women are reported to be older and of smaller stature1-5 and have a higher prevalence of comorbidities.1-6 Studies also show that women present with more advanced symptoms of congestive heart failure despite less extensive coronary artery disease and better preservation of left ventricular function when compared with men.2-4,7 The profile of women undergoing CABG surgery varies among the published investigations because of inconsistency in the data collected by the individual studies and the analysis of the variables.1-12
Variables more commonly prevalent in women were not necessarily predictive of the female patient. Virtually all investigations report that, on average, female patients are significantly older than male patients undergoing CABG surgery. Although we found the prevalence of advanced age to be greater in women, we did not find advanced age to be a significant predictor of female gender in the final parsimonious model.
The preoperative profile for women is distinct from that of the male patient population. Our parsimonious model has high discriminatory ability to distinguish females from males among the demographic, angiographic, and laboratory values from the database. The 18 variables that were statistically predictive of female gender describe the female patient's physiologic "milieu," reflecting the patient's cardiovascular status, organ function, laboratory profile, and physical stature.
Prior studies have described greater disabling symptoms in women despite less extensive coronary artery disease.3,7 We also found that women dominated the New York Heart Association functional classification 3 and 4 categories; however, they had better preserved ejection fractions and less extensive coronary artery disease compared with men. The final parsimonious model did not identify higher New York Heart Association classification categories as a significantly predictive variable for female gender.
In our study, the preoperative lipid profiles for women were distinctly different from men. Women had higher triglyceride, HDL, and LDL cholesterol values. Both the HDL and triglyceride variables were included in the final parsimonious model as significant predictors of female gender. Kannel and Wilson14 mentioned a metabolic link of clustered coronary risk factors among women: dyslipidemia, hypertension, and glucose intolerance. Data from the Framingham Heart Study support that 80% of women with elevated serum cholesterol levels had 1 or more other major risk factors. An incremental increase in the ratio of total and HDL cholesterol "steeply" increased the risk of coronary heart disease–related events.14 Additional laboratory values that distinguished women from men were lower preoperative serum creatinine and bilirubin measurements and lower preoperative red blood cell volume and hematocrit values. Reduced preoperative red blood cell volume measurements in women may have important implications with regard to hemodilution from the cardiopulmonary bypass machine and the subsequent use of red blood cells.
One of the primary limitations of this study is that we can only comment on predictors of female gender based on our measured covariates. Unmeasured variables could impact the final model. Second, our study collected clinical variables and does not explain the purported causes or mechanism for the progression of coronary artery disease. The addition of genetic information to further profile patients may contribute to mechanistic causes in the future. Furthermore, our study was from a single institution and characterizes only those patients with heart disease presenting for coronary artery revascularization.
Gender deserves specific consideration in the broad realm of coronary artery disease. Women may have less risk of developing coronary artery disease earlier in their lives. Data from the Framingham Heart Study indicate that high ratios of total and HDL cholesterol, diabetes, and electrocardiographic evidence of left ventricular hypertrophy eliminate the female advantage.14 The impact of aggressive risk factor modification programs on perioperative outcome is unknown. The key to implementing risk factor modification programs involves an understanding of the high-risk population that the programs are designed to target. A preoperative profile of women such as defined by our model can specifically target patients for intervention.
|
References
|
|---|
- Aldea GS, Gaudiani JM, Shapira OM, Jacobs A, Weinberg J, Cupples A, et al. Effect of gender on postoperative outcomes and hospital stays after coronary artery bypass grafting. Ann Thorac Surg. 1999;67:1097–1103[Abstract/Free Full Text]
- Abramov D, Tamzriz M, Sever J, Christakis G, Bhatnagar G, Heenan A, et al. The influence of gender on the outcome of coronary artery bypass surgery. Ann Thorac Surg. 2000;70:800–806[Abstract/Free Full Text]
- Edwards F, Carey J, Grover F, Bero J, Hartz R. Impact of gender on coronary bypass operative mortality. Ann Thorac Surg. 1998;66:125–131[Abstract/Free Full Text]
- Vaccarino V, Abramson J, Veledar E, Weintraub W. Sex differences in hospital mortality after coronary artery bypass surgery: evidence for a higher mortality in younger women. Circulation. 2002;105:1176–1181[Abstract/Free Full Text]
- Vaccarino V, Lin Z, Kasl S, Mattera, Roumanis S, Abramson J, et al. Gender differences in recovery after coronary artery bypass surgery. J Am Coll Cardiol 2003;41:307–14
- Jacobs A, Kelsey S, Brooks M, Faxon D, Chaitman B, Bittner V, et al. Better outcome for women compared with men undergoing coronary revascularization. A report from the bypass angioplasty revascularization investigation (BARI). Circulation. 1998;98:1279–1285[Abstract/Free Full Text]
- Mickleborough L, Takagi Y, Maruyama H, Sun Z, Mohamed S. Is sex a factor in determining operative risk for aortocoronary bypass graft surgery? Circulation. 1995;92:II-80–84[Medline]
- Brandrup-Wognsen G, Berggren H, Hartford M, Hjalmarson A, Karlsson T, Herlitz J. Female sex is associated with increased mortality and morbidity early, but not late, after coronary artery bypass grafting. Eur Heart J. 1996;17:1426–1431[Abstract/Free Full Text]
- King K, Clark P, Hicks G. Patterns of referral and recovery in women and men undergoing coronary artery bypass grafting. Am J Cardiol. 1992;69:179–182[Medline]
- O'Rourke D, Malenka D, Olmstead E, Quinton H, Sanders J, Lahey S, et al. Improved in-hospital mortality in women undergoing coronary artery bypass grafting. Ann Thorac Surg. 2001;71:507–511[Abstract/Free Full Text]
- Christakis G, Weisel R, Buth K, Fremes S, Rao V, Panagiotopoulos K, et al. Is body size the cause for poor outcomes of coronary artery bypass operations in women? J Thorac Cardiovasc Surg. 1995;110:1344–1358[Abstract/Free Full Text]
- O'Connor G, Morton J, Diehl M, Olmstead E, Coffin L, Levy D, et al. Differences between men and women in hospital mortality associated with coronary artery bypass graft surgery. Circulation. 1993;88:2104–2110[Abstract/Free Full Text]
- Breiman L. Bagging and predicting. Machine Learning. 1996;24:123–140
- Kannel W, Wilson P. Risk factors that attenuate the female coronary disease advantage. Arch Intern Med. 1995;155:57–61[Abstract]
This article has been cited by other articles:
|
|
|
|
 
C. G. Koch
Con: Newly Appreciated Pathophysiology of Ischemic Heart Disease in Women Mandates Changes in Perioperative Management
Anesth. Analg.,
July 1, 2008;
107(1):
33 - 36.
[Full Text]
[PDF]
|
|
|
|
|
|
|
I. K. Toumpoulis, C. E. Anagnostopoulos, S. K. Balaram, C. K. Rokkas, D. G. Swistel, R. C. Ashton Jr, and J. J. DeRose Jr
Assessment of independent predictors for long-term mortality between women and men after coronary artery bypass grafting: Are women different from men?
J. Thorac. Cardiovasc. Surg.,
February 1, 2006;
131(2):
343 - 351.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
F. H. Edwards, V. A. Ferraris, D. M. Shahian, E. Peterson, A. P. Furnary, C. K. Haan, and C. R. Bridges
Gender-Specific Practice Guidelines for Coronary Artery Bypass Surgery: Perioperative Management
Ann. Thorac. Surg.,
June 1, 2005;
79(6):
2189 - 2194.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
F. D. Loop
Coronary Artery Surgery
Ann. Thorac. Surg.,
June 1, 2005;
79(6):
S2221 - S2227.
[Full Text]
[PDF]
|
|
|
Top
Abstract
Methods
