OUP user menu

Impact of body mass index on outcome in patients after coronary artery bypass grafting with and without valve surgery

Evgenij V. Potapov, Matthias Loebe, Stefan Anker, Julia Stein, Selda Bondy, Boris A. Nasseri, Ralf Sodian, Harald Hausmann, Roland Hetzer
DOI: http://dx.doi.org/10.1016/j.ehj.2003.09.005 1933-1941 First published online: 1 November 2003

Abstract

Background Among other preoperative parameters, extremely low or extremely high body mass index (BMI) has been discussed as a substantial risk factor for postoperative complications after cardiac surgery. However, the exact relationship between BMI and postoperative risk has not yet been defined.

Methods We retrospectively investigated consecutive patients (n=22 666) who underwent coronary artery bypass grafting with or without concomitant valve surgery between 1990 and 2001 in our institution. A number of preoperative and intraoperative variables and BMI (as a quadratic term) were used in a logistic regression model as covariates. Further, the patients were divided into 20 groups each with an increase in BMI of 1kg/m2(BMI as a categorical variable). The calculations of odds ratios (ORs) for re-intubation, infection, re-exploration, prolonged stay (>1 day) on the intensive care unit (ICU) and 30-day mortality were adjusted for age, gender and type of surgery.

Results In the multivariate analysis only age (OR between 1.01 and 1.038, P<0.01), additional aortic valve (OR between 1.335 and 2.977, P<0.01) or mitral valve surgery (OR between 2.123 and 3.301, P<0.01) showed significant impact on all five end-points. Patients with BMI between 25 and 35kg/m2were not at elevated risk for any of the investigated end-points, except for infection. Patients with BMI between 21 and 27kg/m2were not at elevated risk for infection. The ORs for postoperative complications were significantly higher in underweight patients compared with obese or severely obese patients, except those for infection. Further, the underweight patients presented significantly more comorbidity.

Conclusion Patients with low BMI are at higher risk after cardiac surgery than obese or severely obese patients. We hypothesize that a preoperative focus on avoiding and/or reversing cachexia may be more efficacious than reducing obesity in reducing the overall risk associated with heart surgery.

  • Bypass grafting
  • BMI
  • Outcome
  • Valve surgery
  • Risk stratification
  • Obesity
  • Cachexia
  • Postoperative complications

1 Background

The body mass index (BMI) expresses nutritional status, metabolic abnormalities and general organ function of patients. A number of studies have shown increased morbidity and mortality in patients with low BMI after cardiac surgery.1,2Obesity is commonly thought to be a risk factor for morbidity and mortality after cardiac surgery. The association of risk factors for coronary artery disease such as hypertension and hypercholesterolaemia with obesity as well as some—often subjective—intra- and postoperative difficulties in obese patients most likely contribute to these perceptions. The data relating to the impact of obesity on postoperative morbidity and mortality are contradictory: many studies showed no significant correlation between increased BMI and mortality after cardiac surgery,1,3–5although in other studies obesity was associated with postoperative infection.1,4–8

We hypothesized that more precise division of patients with regard to BMI would identify patients with optimal BMI and BMI at risk for adverse outcomes and 30-day mortality after cardiac surgery.

2 Methods

Between 1 January 1990 and 31 December 2001, a total of 22 666 consecutive patients underwent coronary artery bypass grafting (CABG) with or without concomitant valve surgery at the Deutsches Herzzentrum Berlin, Germany. All patients were included in this retrospective analysis.

Patients’ data were obtained from the hospital computer data base which uses the standard quality evaluation protocol that is common to all cardiac surgery departments in Germany. Although the database was started in 1986, the present analysis was performed with data from 1990 due to improved quality of the database. The data definitions have not been changed since the protocol was introduced; some new data were added to the protocol but these are not included in the present analysis. The information included the following variables: age, sex, weight, height, diagnosis, priority of the procedure, left ventricular ejection fraction, left ventricular end-diastolic pressure and preoperative medical history, including history of previous cardiac operations, myocardial infarction, diabetes, chronic obstructive pulmonary disease (COPD), hypertension (defined as diastolic blood pressure >95mmHg or use of antihypertension medication), renal insufficiency (serum creatinine >2mg/dl), use of anticoagulants and platelet inhibitors. Recorded operative details included type of surgery (CABG alone, CABG with aortic valve and CABG with mitral valve surgery), cardiopulmonary bypass time, aortic cross-clamp time and operating time. Recorded in-hospital outcomes were re-intubation, re-exploration, infection, prolonged length of stay in the intensive care unit (ICU) and 30-day mortality. Infection was defined as deep sternal or deep leg wound infection requiring additional surgery (superficial leg wound infection was excluded from the analysis), pneumonia, empyema, endocarditis or sepsis. Re-exploration was defined as reopening of the chest during the first 7 days for bleeding or bypass/valve revision. The stay in the ICU was defined as prolonged if it lasted longer than 24h. Priority of surgery was assessed by the surgeons and was defined as follows: ‘elective’ meaning that the operation was indicated but the clinical situation allowed discharge from the hospital with re-admission on the chosen date; ‘urgent’ meaning that the clinical situation required that the patient stay in hospital and the operation be performed before discharge; ‘emergency’ meaning that the clinical situation required immediate operation to prevent morbidity or death.

Operative procedures were performed by a number of surgeons using different surgical techniques. Myocardial protection was performed in all cases using antegrade cold crystalloid cardioplegia. In the early 1990s anesthesia was performed using fentanyl and isoflurane; more recently it has been performed totally intravenously using sufentanyl and propofol.

The BMI was derived from Quetelet’s formula9and equals weight (in kilograms) divided by the square of height (in meters).

3 Statistical analysis

Statistical analysis was performed using SPSS 10.0 for Windows (SPSS Inc. Chicago, IL). The continuous variables are presented as mean±standard deviation. In order to calculate the impact of pre- and intraoperative parameters on the outcomes, a multivariate logistic regression model with stepwise backward elimination procedure was used. The predictors are examined for correlations with each other and included in the logistic regression model only if there are no strong correlations. The correlation coefficients were less than 0.035 for all tested combinations. BMI was included in this procedure as a quadratic term. Due to the large number of patients, the level of significance was set at 0.01.

For assessment of the change of BMI over the time the ANOVA test was used and the year was included as the continuous variable.

For calculation of odds ratios (ORs) for BMI, the patients were divided into 20 groups using BMI as a categorical variable with increments of one point between 18 and 36kg/m2. The small number of patients with BMI <18kg/m2led to the inclusion of these patients in one group. Similarly, patients with BMI ≥36kg/m2were included in one group. For calculation of ORs, for each end-point the group with the lowest risk was identified. This was set as a reference category for calculation of ORs. In accordance with the results of multivariate analysis and data gained in similar studies,5,10the calculations of ORs were adjusted for age, gender and type of surgery in order to eliminate the impact of these parameters. For simplification of the analysis the impact of the fact that in 13% of patients more than one end-point occurred on the stated level of significance for OR was neglected in the calculation of the significance of the OR for different end-points.

For additional analysis, the patients were divided into BMI quintiles. Pre-operative and postoperative parameters in these quintiles were compared using ANOVA for continuous and the χ2-test for categorical variables. Multiple comparisons were adjusted using the Scheffé and Bonferroni-Holm methods respectively. The level of significance was set at 0.01.

4 Results

The distribution of BMI in the study population is shown in Fig. 1. The mean BMI was 26.5±3.82kg/m2, while 0.5% (n=108) of patients presented with BMI <18kg/m2and 1.5% (n=349) of patients had a BMI >36kg/m2. The mean BMI continuously increased over the study period (from 26.1 to 27.2kg/m2, P<0.0001).

Fig. 1

Distribution of BMI in the study population.

The patients in the 4th and 5th quintiles were significantly younger (P<0.0001). Significantly more patients in the 1st quintile were female, had COPD, had a lower ejection fraction and required dialysis. They also required significantly more additional valve procedures with consequently significantly prolonged aortic cross-clamp time. In patients in the first two quintiles the bypass time was significantly prolonged. More underweight patients received urgent or emergency surgery. The patients in the 1st quintile had a higher rate for re-intubation, prolonged ICU stay and mortality. Patients in the 1st and 2nd quintiles had a significantly higher rate for re-exploration. The infection rate was similar in the 1st and 5th quintiles and significantly higher than in the 2nd, 3rd and 4th quintiles. The values in quintiles and significance of the differences between quintiles for perioperative and postoperative data are shown in Tables 1 and 2. In the specific group of patients with BMI <18kg/m2the rate of complications (end-points) was significantly higher than in the total population (re-intubation: 3.7%; re-exploration: 10.3%; infection: 16.7%, prolonged stay in the ICU: 47%; 30-day mortality: 11%).

View this table:
Table 1

Perioperative data of the total study population and of BMI quintiles

ParameterValid casesTotal1st Quintilea2nd Quintile3rd Quintile4th Quintile5th Quintile
BMI (kg/m2)22 59926.5±3.8≤23.523.6–25.325.4–27.127.2–29.4≥29.5
Age (years)22 26863.2±10.363.9±12.1c63.9±10.2c63.5±9.7c62.8±9.3b61.6±9.5a
Weight (kg)22 59976.8±13.561.7±8.3a71.5±6.9b76.4±7.5c82.2±8.1d92.5±12.1e
Height (cm)22 599170.1±8.8168.9±9.2a170.8±8.0b,c170.7±8.2b170.7±8.4b169.4±9.3a,c
Male (%)22 66672.458.4a75.7c77.7c77.9c71.4b
Previous MI (%)21 45254.347.3a55.5b56.2b56.3b55.8b
Previous cardiac surgery (%)22 2428.710.2b9.5a,b8.8a,b8.0a8.0a
LVEF (%)19 25554.31±8.853±16a54±16b55±16b,c55±16b,c55±15c,d
Hypertension (%)16 56568.455.6a64.2b68.1c72.1d80.6e
Diabetes mellitus (%)16 11131.124.7a26.4a29.7b31.5b42.4c
COPD (%)17 37414.817.4d13.9b12.7b13.9a,b,c16.0a,c,d
Preoperative dialysis (%)13 7132.33.9c2.7b,c1.9a,b1.8a,b1.2a
Elective surgery (%)22 66675.573.5a73.9a76.9b,c75.8a,b77.3b,c
Urgent surgery (%)22 66614.515.714.814.314.113.9
Emergency surgery (%)22 6661010.811.38.810.18.8
Patients received CABG alone (%)17 38776.762.3a77.6b80c81c81.6c
CABG with additional AV surgery (%)420218.527.317.816.715.315.8
CABG with additional MV surgery (%)10774.810.44.63.32.92.6
Aortic cross-clamp time (min)22 48650±2353±25b50±24a49±21a50±22a50±21a
Bypass time (min)22 486108.2±62114±69b109±63a,b106±62a107±60a106±54a
  • a Different letters indicate significant differences between quintiles at the level of P<0.01. The same letter indicates no significant differences, for example previous MI (line 6): value for first quintile is significantly lower compared with each other quintile (first quintile marked a, others b). Other quintiles (all marked b): no significant differences. Continuous variables are presented as mean±standard deviation. AV=aortic valve; BMI=body mass index; CABG=coronary artery bypass grafting; COPD=chronic obstructive pulmonary disease; LVEF=left ventricular ejection fraction; MI=myocardial infarction; MV=mitral valve.

View this table:
Table 2

Outcomes in the total study population and of BMI quintiles

ParameterValidcasesTotal1st Quintilea2nd Quintile3rd Quintile4th Quintile5th Quintile
Re-intubation rate (%)22 6653.34.6b3.3a2.5a3.1a3.2a
Re-exploration rate (%)19 3487.19.2b7.8b,c6.6a,c6a5.8a
Infection rate (%)22 66610.311.4c,d9.6a,b8.3a10.3b,c12.2d
Prolonged stay in the ICUb(>1 day) (%)22 42127.433.3e27.4d24.8c24.4b27.6d
Overall 30-day mortality (%)22 6665.17.5c5.6b4.3a4.2a3.5a
  • a Different letters indicate significant differences between quintiles at the level of P<0.01. The same letter indicates no significant differences, for example re-intubation rate (line 1): value for first quintile is significantly higher compared with each other quintile (first quintile marked b, others a). Other quintiles (all marked a): no significant differences.

  • b ICU=intensive care unit.

In the study 65.3% of patients had no complication, 21.7% presented one end-point, 8.4% two, 3.3% three, 1.1% four and 0.1% of patients all five end-points.

Only parameters with significant impact on one or more end-points are presented in Table 3. Two parameters—age and type of surgery—showed a significant impact on all five end-points.

View this table:
Table 3

Results of the backward elimination procedure

ParameterOdds ratios and confidence intervals for
Re-intubationRe-explorationInfectionProlonged ICUastay30-day mortality
Age1.038 (1.029–1.048)1.010 (1.004–1.017)1.034 (1.029–1.040)1.031 (1.027–1.035)1.033 (1.023–1.042)
Male gender1.340 (1.112–1.614)0.812 (0.697–0.946)1.179 (1.088–1.278)
LVEFb0.983 (0.978–0.989)0.978 (0.975–0.998)0.978 (0.976–0.980)0.972 (0.966–0.977)
Previous MIc1.186 (1.028–1.368)1.600 (1.294–1.979)
Previous cardiac surgery1.766 (1.388–2.248)1.480 (1.268–1.728)1.630 (1.452–1.829)2.032 (1.609–2.566)
BMI2d0.999 (0.999–1.000)1.001 (1.001–1.001)1.000 (1.000–1.000)
AVesurgery1.335 (1.076–1.656)1.869 (1.571–2.222)1.455 (1.279–1.655)1.727 (1.575–1.894)2.977 (2.357–3.761)
MVfsurgery2.947 (1.402–2.705)2.342 (1.794–3.056)2.123 (1.717–2.625)2.896 (2.463–3.405)3.301 (2.358–4.623)
Emergency surgery2.212 (1.911–2.562)2.668 (2.380–2.900)3.848 (3.100–4.776)
  • a ICU=intensive care unit.

  • b LVEF=left ventricular ejection fraction.

  • c MI=myocardial infarction.

  • d BMI=body mass index.

  • e AV=aortic valve.

  • f MV=mitral valve.Age, LVEF were analyzed as continuous variables. BMI was analyzed as quadratic term. Odds ratios and 95% confidence intervals (in parentheses) for parameters with significant impact (P<0.01) on analyzed end-points are presented. The odds ratios and confidence intervals for parameters which were eliminated in the backward elimination procedure and therefore not included in the final model are not presented.

The lowest risk for re-intubation was found in patients with BMI between 26 and 26.99kg/m2. In underweight patients (BMI <23kg/m2) and in morbidly obese patients (BMI ≥36kg/m2) the risk was significantly elevated(Fig. 2).

Fig. 2

Relationship between risk for re-intubation in different BMI calculated in relation to the BMI group with lowest incidence (BMI between 26 and 26.99kg/m2). Calculation was adjusted for age, gender and type of surgery. Vertical bars show 95% confidence intervals. *P<0.05; **P<0.01.

The lowest risk for re-exploration occurred in patients with BMI between 27 and 27.99kg/m2. In underweight patients (BMI groups 18–18.99, 20–21.99 and 23–24.99kg/m2) the risk was significantly elevated (Fig. 3).

Fig. 3

Relationship between risk for re-exploration for bleeding in different BMI calculated in relation to the BMI group with lowest incidence (BMI between 27 and 27.99 kg/m2). Calculation was adjusted for age, gender and type of surgery. Vertical bars show 95% confidence intervals. * P<0.05; **P<0.01.

The lowest risk for postoperative infection was presentin patients with BMI between 26 and 26.99kg/m2. In underweight patients (BMI <21kg/m2) and in overweight patients (BMI ≥28kg/m2) the risk was significantly elevated (Fig. 4).

Fig. 4

Relationship between risk for infection in different BMI calculated in relation to the BMI group with lowest incidence (BMI between 26 and 26.99kg/m2). Calculation was adjusted for age, gender and type of surgery. Vertical bars show 95% confidence intervals. *P<0.05; **P<0.01.

The lowest risk for ICU stay longer than 1 day was found in patients with BMI between 27 and 27.99kg/m2. In patients with BMI <25kg/m2and in morbidly obese patients (BMI ≥36kg/m2) the risk was significantly elevated (Fig. 5).

Fig. 5

Relationship between risk for prolonged stay in ICU in different BMI calculated in relation to the BMI group with lowest incidence (BMI between 27 and 27.99kg/m2). Calculation was adjusted for age, gender and type of surgery. Vertical bars show 95% confidence intervals. *P<0.05; **P<0.01.

The lowest risk for 30-day mortality was seen in patients with BMI between 33 and 33.99kg/m2. Only in underweight patients (BMI <22kg/m2) was the risk significantly elevated (Fig. 6).

Fig. 6

Relationship between risk for 30-day mortality in different BMI calculated in relation to the BMI group with lowest incidence (BMI between 33 and 33.99kg/m2). Calculation was adjusted for age, gender and type of surgery. Vertical bars show 95% confidence intervals. *P<0.05; **P<0.01.

These results showed a significantly elevated risk for all analyzed end-points for underweight patients. The severely obese patients did not demonstrate elevated risk for re-exploration or 30-day mortality. The risk for severe infection was significantly elevated in obese patients. The risk for re-intubation and prolonged ICU stay showed a tendency to be elevated in obese patients, but reached significance only in patients with BMI >36kg/m2.

5 Discussion

The major findings of the present study are that overweight patients with a BMI of up to 36kg/m2are at the lowest risk for 30-day mortality after cardiac surgery. Patients with low body weight have a higher risk for evaluated postoperative complications and 30-day mortality than the normal or obese patients. This risk is higher than or similar to that of even severely obese patients. These findings are independent of age, gender and type of surgery.

Recently, several studies have shown elevated mortality and morbidity for cachectic patients after cardiac surgery,1,2,10,11while Gurm et al. suggested more favourable in-hospital outcome in very lean (<20kg/m2) patients.3With regard to the impact of obesity on outcome in cardiac surgery there are different definitions of obesity: some studies of morbidity and mortality defined ‘obesity’ as a BMI >30kg/m2.1,4,6Gurm divided obese patients into two classes (BMI 30 to 34.9 and ≥35kg/m2), while Roques at al. classified patients with a BMI ≥33kg/m2as morbidly obese.3,12Birkmeyer at al.,5on the other hand, defined patients with a BMI ≥36kg/m2as severely obese. The data presented in these studies do not show a uniform relationship between BMI and in-hospital morbidity and mortality.

Moreover, of four accepted risk stratification scores in cardiac surgery (the Euro score, the French score, Parsonnet’s score and the Cleveland Clinic score) not one employs BMI.13,14The patient’s weight is included in Parsonnet’s score as ‘morbid obesity’ defined by weight as a risk factor and in the Cleveland Clinic score as one point for weight <65kg. A possible explanation for this fact is the elimination of the BMI in the multivariate logistic regression models if analyzed as a continuous variable10because of the U- or J-shaped relationship between BMI and risk for adverse outcome (Figs. 2–6). In our study, when the BMI was analyzed as a continuous variable (Table 3), it showed only low impact on three of the five analyzed end-points. This suggests that analysis of BMI as a continuous variable is inappropriate for risk stratification in cardiac surgery. However, another explanation would be a correlation between gender and BMI.2,10

The large number of patients in our study allowed their division into 20 groups for calculation of the odds ratios. Adjustment of the analysis for age, gender and type of surgery was performed based on the results of multivariate analysis, which showed that age and type of surgery are independent predictors for all investigated end-points of the study. Other investigators also showed significant impact of age2,5,10and gender2,10on operative mortality. Our approach led to better analysis of marginal groups (BMI <20 and ≥30kg/m2) and showed that the group of ‘obese’ patients is not homogenous with regard to risk for adverse outcome. The range of BMI with non-elevated risk for the investigated end-points, except infection, included overweight and obese patients (Figs. 2, 3, 5 and 6). Patients without risk for infection have a BMI between 21 and 27kg/m2. Overweight patients were at elevated risk for postoperative infection (Fig. 4) and severely obese patients (BMI of 36kg/m2and over) were at elevated risk for prolonged ICU stay and re-intubation, while the underweight patients have elevated risk for all investigated end-points. This fact is strongly supported by the finding that cachectic patients (BMI <18kg/m2) presented significantly more complications compared with the overall study population and, except for re-intubation, with all five quintiles. These facts suggest that for stratification of perioperative risk the patients with low body weight and especially cachectic patients should always be taken into consideration, but ‘severely obese’ patients only for re-intubation, prolonged ICU stay and infection. This supports the findings of previous studies1,4–8that the risk for postoperative infection, especially for deep sternal infection, is significantly elevated in severely obese patients. The decreased perfusion of subcutaneous fat tissue and mechanical tension may lead to this phenomenon. In these patients efforts directed at perioperative infection prophylaxis employing more aggressive use of antibiotics in the postoperative period, partial sternotomy or use of enhanced techniques for sternal closure15–17could be helpful in reducing sternal infection. Less invasive vein harvesting techniques18may reduce leg wound infection. Furthermore, more aggressive treatment of hyperglycemia in patients with diabetes mellitus in the postoperative period may also reduce wound infection in obese patients. It is likely that a large prospective study would show the benefit.

As yet, no studies could define the origins of elevated morbidity and mortality in patients with low BMI undergoing cardiac operations. Comorbidity, gender, small coronary vessels, implantation of smaller valves, and increased hemodilution have been discussed, but no significant conclusions could be drawn.1,2,11The presence of severe noncardiovascular underlying diseases in lean patients, which has been shown in epidemiologicaltrials,19,20may contribute to this effect. In our study the underweight patients also presented more comorbidity such as COPD and greater need for preoperative dialysis and additional valve surgery. Moreover, the patients in the 1st and 2nd quintiles have more acute cardiac events leading to significantly more urgent or emergency operations (Table 1). However, the low BMI may be a surrogate marker for comorbidities.

A recent study1showed no association between BMI and nutritional status, as expressed by the serum albumin level,21with regard to morbidity and mortality in patients undergoing cardiac surgery. However, low serum albumin levels (<2.5g/dl) are more frequent in patients with low body weight compared with obese patients (21 vs 11%).1This suggests that patients with a higher percentage of body fat may have better nutritional status and therefore more reserves, which may allow them to handle operative stress and postoperative complications more efficiently. Nutritional status may be an important factor with impact on morbidity and mortality after cardiac surgery.

6 Limitations of the study

The study represents our single center experience over a period of 12 years. The changes in technique and patient population over time, such as the increased number of older or obese patients, may have an impact on the results despite age adjustment of the calculation of the odds ratios.

A BMI-based definition describes the relation between weight and height and fails to take into account body fat distribution, which may be a better predictor for cardiovascular risk.

Both cardiologists and surgeons may have been biased against multimorbid obese patients whereas they may have been less reluctant to accept underweight patients for surgery. Be that as it may, the patient selection in this study can be considered to be representative for the current approach to obese patients and patients with low body weight.

The albumin levels were not available for statistical analysis and this fact precludes the study of nutritional status in the study population. Further, incomplete information about the size of the coronary arteries and of long-term follow up prevents the inclusion of these parameters in the statistical analysis.

7 Conclusion

After elimination of the impact of age, gender and type of surgery, patients with low BMI are seen to be at higher risk for postoperative complications after cardiac surgery than normal or even severely obese patients, while obesity is not a risk factor for adverse outcome except for infection. Low body weight (BMI <20kg/m2) should be considered as a risk factor in preoperative risk stratification scores in cardiac surgery. However, BMI ≥36kg/m2is a risk factor for postoperative infection. Pre-operative efforts to normalize nutritional status and increase weight in patients with low body weight may have more effect on postoperative morbidity and mortality than weight reduction in obese and severely obese patients.

Acknowledgments

We are grateful to Detlef Gösmann and Michelle Menzel for help in data collection and analysis of the huge database and to Anne Gale for editorial assistance.

Footnotes

  • 1 Presented in part at the Annual Meeting of the European Society of Cardiology, Berlin, Germany, 4 September 2002

References

View Abstract