The obesity paradox is still there: a risk analysis of over cardiosurgical patients based on body mass index

Transcription

1 Interactive CardioVascular and Thoracic Surgery 25 (2017) doi: /icvts/ivx058 Advance Access publication 18 March 2017 ORIGINAL ARTICLE Cite this article as: Hartrumpf M, Kuehnel R-U, Albes JM. The obesity paradox is still there: a risk analysis of over cardiosurgical patients based on body mass index. Interact CardioVasc Thorac Surg 2017;25: The obesity paradox is still there: a risk analysis of over cardiosurgical patients based on body mass index Martin Hartrumpf*, Ralf-Uwe Kuehnel and Johannes M. Albes Department of Cardiovascular Surgery, Heart Centre Brandenburg, University Hospital, Medical School Brandenburg, Bernau, Germany * Corresponding author. Department of Cardiovascular Surgery, Heart Centre Brandenburg, Ladeburger Strasse 17, Bernau, Germany. Tel: ; fax: ; m.hartrumpf@immanuel.de (M. Hartrumpf). Received 18 July 2016; received in revised form 15 December 2016; accepted 21 January 2017 Abstract OBJECTIVES: Obesity is an ever-growing problem in contemporary cardiac surgery. Although it accounts for many perioperative comorbidities, it has not been shown to increase mortality. Body mass index (BMI) is therefore not considered in the European System for Cardiac Operative Risk Evaluation (EuroSCORE). We sought to confirm whether this holds true for our own single-centre patient population. METHODS: Data from consecutive patients receiving major cardiac surgery at our institution were analysed. Gender, age, BMI, EuroSCORE, urgency, redo status and all-cause in-hospital mortality were derived from our database. Mortality was grouped into 4 BMI categories. We created a logistic regression model to identify predictors of mortality. RESULTS: There were males and 4280 females. Categorical mortality was 8.79% (underweight), 7.04% (normal weight), 5.16% (overweight), 6.30% (obese), rendering an inverse J-shaped pattern known as obesity paradox. Univariable regression detected significant predictors of mortality: rising age, female gender, urgent procedures, redo surgery (P < 0.001). BMI was no predictor (P = 0.575) but became significant with the multivariable analysis (P = 0.004). Its effect on mortality was exclusively indirect, being mediated through age (P < 0.001). Receiver-operating characteristics curve analysis also confirmed that BMI did not qualify as a risk factor. However, the overweight category was a predictor of lower mortality. CONCLUSIONS: Our findings from > patients confirm the obesity paradox showing the least mortality in the overweight group. They support the current EuroSCORE model in that BMI is no independent predictor of early mortality. However, such patients still carry the risk of comorbidities. Likewise, special care is required with underweight patients who show the highest in-hospital mortality. Keywords: Cardiac surgery Body mass index Risk factors Mortality INTRODUCTION Overweight and obesity represent a worldwide health problem that is no longer restricted to developed countries. According to current WHO data, >1.9 billion adults were overweight in 2014 and over 600 million of these were obese. The numbers have doubled since 1980 [1]. Quite a similar situation appears in the German population, where overweight has been constant at a high level, but obesity has dramatically grown over the last 2 decades. This is particularly true for adolescents and young adults. Male adults are more affected than females. The mean body mass index (BMI) across all ages was 26.7 kg/m 2 for men and 25.4 kg/m 2 for women in the last German National Health Survey (DEGS1) [2]. Obesity is a known risk factor for cardiovascular diseases, such as heart disease and stroke, hypertension, hyperlipidaemia, diabetes, musculoskeletal disorders and some malignancies (endometrial, breast and colon cancer) [1, 3, 4]. It accounts for a higher worldwide mortality than underweight [1]. Its prevalence is equally reflected in the current cardiosurgical patient population. Common sense would expect an elevated mortality in obese patients who undergo cardiac surgery. As known from daily experience, this subpopulation is more likely to present with adverse conditions, such as complex anatomy, deep surgical situs, prolonged intensive care unit stay, delicate ventilation, delayed weaning, tracheostomy, pulmonary infections, diabetes, impaired wound healing, sternal dehiscence, orthopaedic problems or immobilization [5, 6]. Bleeding does not seem to be an issue, as it occurs less frequently than in non-overweight patients [5 7]. Surprisingly, obesity has not been identified as a risk factor for cardiosurgical mortality in many contemporary studies [5 12]. Thus, it has never entered the European System for Cardiac Operative Risk Evaluation (EuroSCORE) risk model as an independent predictor [13, 14]. Quite the contrary, moderately overweight patients show reduced cardiosurgical and general cardiovascular mortality referred to as obesity paradox [3, 15 19]. Evidence exists that the logistic EuroSCORE may be inaccurate due to missing variables such as BMI VC The Author Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.

2 M. Hartrumpf et al. / Interactive CardioVascular and Thoracic Surgery 19 [20]. However, the relevance of a simple BMI measurement to reflect the actual health condition remains controversial [3]. In the light of this, we sought to clarify from our single-centre experience with > patients whether or not the BMI has an impact on short-term operative mortality. METHODS Study design This study was designed as a retrospective, observational, all-comers study. Data were taken from our institution s patient database over the full available time range according to inclusion criteria. Primary end-point was all-cause in-hospital mortality. Aim of the study was to identify independent risk factors of mortality and to relate different factors to patients BMI. Data collection Our clinical database was retrospectively analysed from January 2004 to July During this period, all consecutive patients receiving major cardiac surgery were included. These comprised all cardiosurgical patients operated on cardiopulmonary bypass (bypass and valve surgery, aortic surgery, cardiac tumours, congenital defects and pericardial diseases) as well as major off-pump surgery (minimally invasive direct coronary artery bypass, offpump coronary artery bypass and transapical transcatheter aortic valve implantation). Minor off-pump surgery, such as pacemaker implantation or pericardial drainage, was not included. Also excluded were non-cardiac operations, such as lung resections, even when supported on cardiopulmonary bypass. Patient gender, age, height, weight, BMI, logistic EuroSCORE, status of the operation, redo status and all-cause in-hospital mortality were derived from the database (logistic EuroSCORE available only since July 2006). BMI was categorized according to WHO standard as follows [21]: Underweight <18.50 kg/m 2 Normal weight kg/m 2 Overweight kg/m 2 Obese >_30.00 kg/m 2 As for the status of the operation, we used the following common definitions as applied in the context of EuroSCORE [22]: Elective: regularly scheduled Urgent: next working day Emergency: immediately (same day) Ultima ratio: immediately, at highest risk, salvage Surgical technique On-pump surgery was usually performed using ascending aortic cannulation with normothermia or mild hypothermia (28 to 34 C). Venous drainage was accomplished either with right atrial or with bicaval cannulation as required. During aortic arch surgery, deep hypothermic circulatory arrest (<20 C) and antegrade cerebral perfusion were established and retrograde femoral or antegrade (aortic, axillary) perfusion was used according to clinical requirement. Standard access was a median sternotomy. For minimally invasive procedures, partial sternotomy or limited anterolateral thoracotomy were also used. All patients received general anaesthesia and were extubated on the intensive care unit as early as possible. Intraoperative transoesophageal echocardiography was routinely used in all valve procedures. Statistical methods Data analysis was performed using SPSS Version 17.0 (SPSS Inc., Chicago, IL, USA). Descriptive data are expressed as mean ± standard deviation or percentages, respectively. Gender differences and differences between surviving and deceased patients were assessed using unpaired t-test for continuous variables or v 2 test for categorical variables. For detection of dependencies between numerical variables, correlation analysis and linear curve fitting were applied and scatter plots were generated. Spearman s correlation coefficients and their squares were calculated. Univariable analysis of variance was performed to relate logistic EuroSCORE to the different BMI categories followed by Tukey s post hoc test to correct for multiple comparisons. Distribution of the different surgical procedures across the BMI categories was assessed by creating 5 categories according to the key procedure. To assess the risk factors of mortality, we created a binary logistic regression model using the following independent variables: age and BMI (both continuous), gender, urgency status and redo status (categorical). Urgency categories were condensed into no emergency (elective, urgent) and emergency (emergency, ultima ratio) to deflate the model. Dummy coding for the categorical variables was as follows: male = 0, female = 1; no emergency = 0, emergency = 1; and nonredo = 0, redo = 1. Logistic EuroSCORE was not included as a variable to avoid multicollinearity, as it would contain redundant information about age, gender, urgency and redo status. An additional logistic regression model was created using categorical instead of numerical BMI. Odds ratios with their confidence intervals and P-values were calculated. Receiver-operating characteristics curves for the continuous variables were calculated to determine accuracy of prediction. Mediation analysis [23] was performed using the free PROCESS plugin provided by Andrew F. Hayes ( to detect suspected interrelations of the independent variables. RESULTS A total of patients were included in the study (72.1% male). Men showed a greater body height and weight and a slightly different urgency pattern than women. Redo status showed no gender difference. Women were older (70.3 ± 9.5 vs 66.7 ± 10.0 years, P < 0.001), had a higher BMI (28.77 ± 5.59 vs ± 4.43 kg/m 2, P < 0.001) and were at higher surgical risk (logistic EuroSCORE, ± vs ± 14.63%, P < 0.001). This resulted in a higher overall mortality compared with men (8.1% vs 5.2%, P < 0.001). Mortality was higher in patients undergoing redo surgery compared with first surgery (17.7% vs 5.4%, P < 0.001). Urgency correlated with mortality in ascending order (elective 3.5%, urgent 6.4%, emergency 15.3% and ultima ratio 50.0%, P < 0.001). Median postoperative length of stay was 13 days. Table 1 informs about BMI categorization. Men tended to overweight rather than obesity, while women showed balanced fractions here. More female than male patients were at normal weight. Table 2 shows the comparison between surviving and ADULT CARDIAC

3 20 M. Hartrumpf et al. / Interactive CardioVascular and Thoracic Surgery deceased patients. Of note, there was no difference regarding the numerical BMI. Regression analysis with linear curve fitting showed that logistic EuroSCORE significantly increased with age (Fig. 1C) as age is included in the score. Correlation between age/weight was weak (Fig. 1A) while there was practically no correlation between age/ BMI or BMI/EuroSCORE, respectively (Fig. 1B and D). When considering BMI categories, univariable analysis of variance shows that the significant EuroSCORE differences are only situated across the 2 lower versus the 2 upper BMI categories (Fig. 2A). In Fig. 2B, it is demonstrated that female patients are equally aged across the different BMI categories, while male patients are somewhat younger in the extreme categories (i.e. underweight and obese) compared with the centre categories. Figure 2C shows a reverse J-shaped distribution of observed mortality across the BMI categories with a minimum for the overweight patients (BMI kg/m 2 ) and a maximum for the underweight patients (BMI <18.50 kg/m 2 ), Table 1: Distribution of patients across BMI categories Under weight Normal weight Over weight Obese n Median age Male gender, % % of Males, % % of Females, % Shown are the median age and the fraction of males for each category and the gender distribution across the different categories. P < Figure 3 shows the percentage distribution of operative procedures across the different BMI categories. To assess the risk factors of in-hospital mortality, 2 alternative models of logistic regression were created, using either numerical or categorical BMI (Table 3). In summary, univariable and multivariable logistic regression revealed that advancing age, female gender, urgency and redo status were highly significant risk factors for increased in-hospital mortality. In contrast, numerical BMI did not come up as a risk factor in the univariable analysis (P = 0.575) but, however, became significant in the presence of the age variable in the multivariable analysis (P = 0.004). There was no direct interaction between age and BMI (P = 0.793) or gender and BMI (P = 0.234). After replacing numerical BMI with categorical BMI (using normal weight as reference), BMI became a significant predictor. In detail, only overweight patients showed a reduced risk of mortality while underweight and obese patients did not. Again, there was no interaction of age or gender with BMI. Table 3 shows the results for the univariable and both multivariable analyses. As for multivariable regression, of the total cases entered the analysis due to some missingvalues.intotal,therewere 917 mortalities (6.0%). Using stepwise forward inclusion, all 5 independent variables entered the model while showing significant improvement with each step (omnibus test, P <0.001 each). The final regression model showed a borderline overall adaptation: -2LL = 6229, Cox and Snell R 2 =0.045, Nagelkerke R 2 = Hosmer Lemeshow test rendered a v 2 of (P =0.840),thusindicating acceptable predictive power of the model. However, while 99.9% of survivors ( of ) were correctly predicted, prediction of mortality still remained poor (20 of 917 = 2.2%). Receiver-operating characteristics curve analysis with respect to mortality (Fig. 4) revealed that the predictive power of age is relevant (area under the curve = 0.653, 95% confidence interval = to 0.671) while the predictive power of BMI is close to Table 2: Characteristics of surviving and deceased patients Total Alive Deceased P-value Female gender, % <0.001 Age, years 67.5 ± ± ± 8.7 <0.001 Body height, cm ± ± ± 9.1 <0.001 Body weight, kg 82.5 ± ± ± BMI, kg/m ± ± ± ES (add), points 6.3 ± ± ± 4.5 <0.001 ES (log), % ± ± ± <0.001 Status of operation, % Elective <0.001 Urgent Emergency Ultima ratio BMI (cat), % Underweight Normal Overweight Obese Redo, % <0.001 In-hospital mortality (n = 910), % 6.0 Data are displayed as mean ± standard deviation (SD), or percentages, respectively. Shown are the values for the total population as well as separated into surviving and deceased patients. P-values refer to alive versus deceased. BMI: body mass index; BMI (cat): categorical body mass index; ES (add): additive EuroSCORE; ES (log): logistic EuroSCORE.

4 M. Hartrumpf et al. / Interactive CardioVascular and Thoracic Surgery 21 ADULT CARDIAC Figure 1: Correlation analysis and linear curve fitting. (A) body weight vs age; (B) BMI vs age; (C) logistic EuroSCORE vs age; (D) logistic EuroSCORE vs BMI. chance (area under the curve = 0.495, 95% confidence interval = to 0.516). DISCUSSION First published in 1999 [13], the EuroSCORE has become a powerful tool for cardiosurgical risk prediction in the European countries. It was also validated for the North American population in 2002 [24]. The original EuroSCORE was based on a logistic regression model using 17 items for risk calculation, not including BMI. However, evidence exists that it may be inaccurate in lowrisk patients due to missing variables such as BMI, diabetes, creatine kinase myocardial isoenzyme, and left internal mammary artery usage, which also affect long-term survival [20]. During the last 2 decades, cardiosurgical mortality has decreased although the patients have become older and sicker. Due to some shortcomings in accuracy and overprediction of mortality, an updated version of the risk model (EuroSCORE II) was published in 2012 [14]. It predicts mortality more precisely and is still the current model for today s use. But again, BMI failed to show up as a predictor of cardiosurgical mortality. From daily clinical experience, it is known that obese patients frequently present with higher postoperative morbidity. Reasons for this include delayed weaning from the respirator, pulmonary infections, metabolic disorders, impaired mobilization, reduced sternal wound healing or local infections. Not uncommon that the question arises whether patients with high BMI do suffer an increased mortality as well [20]. Current literature shows several studies focusing on the impact of BMI on mortality in different cardiosurgical patient populations. The most recent studies were performed in North America [11, 19], South America [5], in Arabic [6, 7] and East European countries [9, 10] and Australia [25]. There is a small follow-up study from the Netherlands [26]. To the best of our knowledge, there has been no extensive analysis of this particular question in a considerable West European cardiosurgical population independent of the EuroSCORE database. Therefore, we retrospectively assessed > unselected consecutive patients undergoing major cardiosurgical procedures. Our cohort showed the typical composition of an ageing patient population. There was a predominance of male patients (72.1%), but women were considerably older, showing a higher operative risk and thus a significantly higher mortality. It should be kept in mind that the in-hospital mortality is dependent on the institution s transfer policy. Unlike other hospitals, we do not routinely transfer patients early after surgery. Our patients showed a higher average BMI than the entire German population, where the total BMI was reported 25.4 kg/m 2 for women and 26.7 kg/m 2 for men [2]. After categorizing BMI [21], it could be seen that the principal risk transition occurred between normal weight and overweight because overweight and obese patients presented with the smallest average risk scores (Fig. 2A). The vast majority of our patients were overweight showing a predominance of males. The second largest group were the obese who were slightly younger and showed a lower rate of males. In contrast, the underweight group showed an intermediate age and equally contained men and women. Different results have been reported from a large Canadian cardiosurgical population [17]. Here the oldest patients were found in the underweight group (67.81 ± years), followed by normal weight (66.86 ± 10.23), overweight (65.37 ± 9.90), obesity (63.67 ± 9.73) and morbid obesity

5 22 M. Hartrumpf et al. / Interactive CardioVascular and Thoracic Surgery Figure 3: Distribution of operative procedures across the different BMI categories ("case mix"). Table 3: Univariable and multivariable results of logistic regression analysis OR 95% CI P-value Lower Upper Univariable analysis Age <0.001 Gender <0.001 BMI BMI (cat) Underweight Overweight <0.001 Obese Urgent <0.001 Redo <0.001 Multivariable analysis BMI continuous Age <0.001 Gender <0.001 BMI Urgent <0.001 Redo <0.001 Multivariable analysis BMI categorical Age <0.001 Gender <0.001 BMI (cat) Underweight Overweight Obese Urgent <0.001 Redo <0.001 Figure 2: (A) Mean logistic EuroSCORE related to the BMI categories. Significant differences between groups are indicated (Tukey s post hoc test); (B) age distribution over the BMI categories with regard to gender; and (C) mortality related to the different BMI categories. (61.78 ± 9.61). It must be taken into account that the underweight fraction was very low (<2%) in both studies which renders results less accurate due to a higher standard error. Similar to our study, the overweight group showed the highest fraction of men (82.2%), while the underweight group was more Continuous variables are age [years] and BMI [kg/m 2 ]. Categorical variables are BMI (cat), gender [0 = male, 1 = female], urgency status [0 = no emergency, 1 = emergency] and redo status [0 = non-redo, 1 = redo surgery]. BMI categories are referenced to normal weight. BMI: body mass index; BMI (cat): categorical body mass index; CI: confidence interval for odds ratio; OR: odds ratio. balanced (59.2%). Another North American study showed that the most obese patients were significantly younger and more likely to be female [11]. Subgroup analysis of our underweight patients showed that there were many elderly women likely to be frail, whereas slender men were significantly younger,

6 M. Hartrumpf et al. / Interactive CardioVascular and Thoracic Surgery 23 Figure 5: Result of mediation analysis using the SPSS PROCESS Plugin, see Ref. [23]. Shown is the linear relationship between BMI, age and mortality. ADULT CARDIAC Figure 4: Receiver-operating characteristics curve analysis for age (straight curve) and BMI (dotted curve) with respect to predicted mortality. including adolescents. No underweight patient who died was young. A remarkable finding that explains the lack of significance for numerical BMI in the logistic regression model is the non-linear behaviour of mortality across the BMI classes (Fig. 2C). We could show that the underweight group by far showed the highest mortality. This may be related to its relevant proportion of frail women. In contrast, slightly overweight patients exhibited the lowest mortality around 5%, while those with either normal weight or obesity showed a somewhat higher mortality. This behaviour is referred to as reverse J-shaped or U-shaped and reflects the fact that male lower risk patients are over-represented in the overweight group. Very similar results were reported from Canadian patients [17]. Here, hazard ratios were calculated against normal weight as follows: underweight = 1.35, overweight = 0.94, obese = 1.00 and morbidly obese = The authors derived the same reverse J-shaped relationship [17]. In our population, mortality in obese patients was even lower than in those with normal weight. Several studies confirm our findings. The data set by van Straten et al. [26] focused on 1758 isolated aortic valve replacement patients and failed to show an association between early mortality and BMI. Follow-up confirmed that only underweight, but not obesity, was an independent predictor of late mortality. Another study from Turkey in 803 coronary artery bypass grafting patients demonstrated that underweight is an independent risk factor of early mortality and postoperative complications in contrast to normal or overweight patients [10]. A large retrospective study from Atlanta including 4247 valvular surgery patients showed that moderate overweight (BMI kg/m 2 ) resulted in the shortest length of stay and lowest short-term mortality, while patients with lower BMI suffered the highest short- and long-term mortality. However, the cut-off was set at 24 kg/m 2, thus including normal weight patients [11]. Interestingly, an older study from Germany actively determined the same cut-off (24 kg/m 2 ), which divided the 1241 patients in such a way that the dichotomous model became equivalent to the linear BMI model [12]. Our observed risk transition (Fig. 2) corresponds to this cut-off value very well. With regard to Fig. 2C, we believed that introducing the categorical BMI into the logistic regression model would better adapt to reality. Results were similar except that overweight patients were at significantly lower risk of mortality compared with normal weight patients, which rendered the BMI variable significant. In this, our results statistically confirm the obesity paradox [3, 15 18], which describes that moderately overweight patients show reduced cardiosurgical and general cardiovascular morbidity and mortality. This holds true unless someone shows extreme obesity (BMI >40 kg/m 2 ) [19]. Underweight people may be at higher risk as they are often older and more likely suffer from malignancies, heart failure, malnutrition and organ dysfunction than normal-to-overweight people. Gradual weight loss may occur, and if considered, the paradox disappears [27]. Obese patients are often encountered earlier in the health care system than underweight patients, and there might be stricter adherence to the guidelines [16]. Our data are not sufficient to support or disconfirm these theories. However, we could show that part of this paradox could be explained by the patient age, which mediated the effect of BMI on mortality (see below). Moreover, there were slight differences in the distribution of operative procedures across the different BMI categories that may have influenced mortality. This has also been observed by other authors [19]. Our results show that in the overweight and obese patients, high-risk open-heart procedures (valves, combined procedures and aortic surgery) were less frequently performed than in low-weight patients, which may reflect their lower mortality. Meanwhile there is some general doubt about the obesity paradox. Clinical data from the real world do not consistently support this theory, and thus, there is no strict evidence for such a paradox [16, 27]. The clinical course of a patient is much dependent on metabolic conditions that are not yet fully understood so that BMI measurement alone will not sufficiently explain the paradox. In the light of this, active weight gain cannot be recommended [3]. Moreover, BMI does not account for body composition or the location of adipose tissue, which may change with age [17]. In our study, we could not confirm a BMI difference between patients who died early and those who were discharged from the hospital. Likewise, numerical BMI was no independent predictor of mortality in the univariable logistic regression. Mortality was rather related to advanced age, female gender, emergencies and redo surgery. This is consistent with other studies [7]. In the multivariable logistic regression, however, BMI became a significant predictor in the presence of age. Interaction analysis showed that BMI did not have a direct effect on mortality but that its effect was rather mediated through patient age (Fig. 5). In fact, the higher the BMI, the lower the age, the lower the mortality. This leads to an indirect negative effect which is very small (coefficient

7 24 M. Hartrumpf et al. / Interactive CardioVascular and Thoracic Surgery close to 0) but renders the variable significant in the overall analysis. When categorical BMI was used instead, overweight became a significant predictor of reduced mortality, but obesity or underweight showed no different risk than normal weight. Finally, receiver-operating characteristics curve analysis confirmed that mortality was not accurately predicted by BMI itself. Taken together, BMI did not qualify as a risk factor of early mortality in our patient population. Limitations This study is observational in nature and limited by the restricted availability of clinical data in our digital database. There was no detailed information about patient history (e.g. smoking or vascular disease) or any follow-up beyond hospital discharge. Thus, no subgroup assessment or survival analysis could be performed. Only all-cause in-hospital mortality could be determined, as we had no digital information about specific causes of death. Inhospital mortality depends on a hospital s patient transfer policy, so it might not be well comparable to other institutions. However, even though the bandwidth was limited, all data actually used for the study were reliably collected in all patients. CONCLUSIONS In conclusion, our findings support the EuroSCORE model where BMI has never been included as a risk factor. From our singlecentre experience with > patients, we could show that BMI is not an independent predictor of early cardiosurgical mortality as opposed to age, gender, urgency and redo status. However, care should still be taken since obese patients often suffer from various comorbidities that may adversely affect their clinical course [17]. Special focus must be placed on underweight patients as these showed the highest mortality and may have the least reserves. This holds particularly true for elderly patients who are likely to be frail. Cautious risk assessment, preconditioning and careful allocation to surgery are therefore mandatory. Conflict of interest: none declared. REFERENCES [1] World Health Organization (WHO). Obesity and Overweight: Fact sheet No (updated June 2016, 4 July 2016, date last accessed). [2] Mensink GB, Schienkiewitz A, Haftenberger M, Lampert T, Ziese T, Scheidt-Nave C. Overweight and obesity in Germany: results of the German Health Interview and Examination Survey for Adults (DEGS1). Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2013;56: [3] Goyal A, Nimmakayala KR, Zonszein J. Is there a paradox in obesity? Cardiol Rev 2014;22: [4] Whitlock G, Lewington S, Sherliker P, Clarke R, Emberson J, Halsey J et al. Body-mass index and cause-specific mortality in adults: collaborative analyses of 57 prospective studies. Lancet 2009;373: [5] Costa VE, Ferolla SM, Reis TO, Rabello RR, Rocha EA, Couto CM et al. Impact of body mass index on outcome in patients undergoing coronary artery bypass grafting and/or valve replacement surgery. Rev Bras Cir Cardiovasc 2015;30: [6] Allama A, Ibrahim I, Abdallah A, Ashraf S, Youhana A, Kumar P et al. Effect of body mass index on early clinical outcomes after cardiac surgery. Asian Cardiovasc Thorac Ann 2013;22: [7] Shahabuddin S, Perveen S, Furnaz S, Fatimi S, Sami S, Sharif H. Body mass index predictor of outcome after coronary artery bypass grafting. Asian Cardiovasc Thorac Ann 2013;21: [8] Jörg FRG. Stellt Adipositas einen isolierten Risikofaktor in der Kardiovaskularchirurgie dar? Eine retrospektive Untersuchung an 2251 Patienten mit Body Mass Index >30 und spezieller Betrachtung bilateraler IMA-Grafts. Thesis in German. Technische Universit at München (Published 10 April 2013, 4 July 2016, date last accessed). [9] Cemerlic-Adjic N, Pavlovic K, Jevtic M, Velicki R, Kostovski S, Velicki L. The impact of obesity on early mortality after coronary artery bypass grafting. Vojnosanit Pregl 2014;71: [10] Atalan N, Fazliogullari O, Kunt AT, Basaran C, Gurer O, Sitilci T et al. Effect of body mass index on early morbidity and mortality after isolated coronary artery bypass graft surgery. J Cardiothorac Vasc Anesth 2012;26: [11] Thourani VH, Keeling WB, Kilgo PD, Puskas JD, Lattouf OM, Chen EP et al. The impact of body mass index on morbidity and short- and longterm mortality in cardiac valvular surgery. J Thorac Cardiovasc Surg 2011;142: [12] Florath I, Albert AA, Rosendahl UP, Hassanein WM, Bauer S, Ennker IC et al. Body mass index: a risk factor for 30-day or six-month mortality in patients undergoing aortic valve replacement? J Heart Valve Dis 2006;15: [13] Nashef SA, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999;16:9 13. [14] Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Goldstone AR et al. EuroSCORE II. Eur J Cardiothorac Surg 2012;41: [15] Takagi H, Umemoto T. Overweight, but not obesity, paradox on mortality following coronary artery bypass grafting. J Cardiol 2016;68: [16] Akin I, Nienaber CA. Obesity paradox in coronary artery disease. World J Cardiol 2015;7: [17] Johnson AP, Parlow JL, Whitehead M, Xu J, Rohland S, Milne B. Body mass index, outcomes, and mortality following cardiac surgery in Ontario, Canada. J Am Heart Assoc 2015;4:1 13. [18] Pickkers P, de Keizer N, Dusseljee J, Weerheijm D, van der Hoeven JG, Peek N. Body mass index is associated with hospital mortality in critically ill patients: an observational cohort study. Crit Care Med 2013;41: [19] Gao M, Sun J, Young N, Boyd D, Atkins Z, Li Z et al. Impact of body mass index on outcomes in cardiac surgery. J Cardiothorac Vasc Anesth 2016;30: [20] O Boyle F, Mediratta N, Fabri B, Pullan M, Chalmers J, McShane J et al. Long-term survival after coronary artery bypass surgery stratified by EuroSCORE. Eur J Cardiothorac Surg 2012;42: [21] World Health Organization (WHO). Global Database on Body Mass Index BMI Classification. intropage=intro_3.html (4 July 2016, date last accessed). [22] Roques F, Nashef SA, Michel P, Gauducheau E, de Vincentiis C, Baudet E et al. Risk factors and outcome in European cardiac surgery: analysis of the EuroSCORE multinational database of patients. Eur J Cardiothorac Surg 1999;15: [23] Andy Field. Discovering Statistics Using IBM SPSS Statistics. London: SAGE Publications Ltd., 2013, [24] Nashef SA, Roques F, Hammill BG, Peterson ED, Michel P, Grover FL et al. Validation of European System for Cardiac Operative Risk Evaluation (EuroSCORE) in North American cardiac surgery. Eur J Cardiothorac Surg 2002;22: [25] Ho KM, Bertenshaw C, Same S, Schneider M, Williams KA, Godsell T et al. Differential associations between body mass index and outcomes after elective adult cardiac surgery: a linked data cohort study. Anaesth Intensive Care 2013;41: [26] van Straten AH, Safari M, Ozdemir HI, Elenbaas TW, Hamad MA. Does the body mass index predict mortality after isolated aortic valve replacement? J Heart Valve Dis 2013;22: [27] Stokes A, Preston SH. Smoking and reverse causation create an obesity paradox in cardiovascular disease. Obesity (Silver Spring) 2015;23: