Access the full text.
Sign up today, get DeepDyve free for 14 days.
Candyce Kroenke, Wendy Chen, B. Rosner, M. Holmes (2005)
Weight, weight gain, and survival after breast cancer diagnosis.Journal of clinical oncology : official journal of the American Society of Clinical Oncology, 23 7
K. Bhaskaran, I. dos-Santos-Silva, D. Leon, I. Douglas, L. Smeeth (2018)
Association of BMI with overall and cause-specific mortality: a population-based cohort study of 3·6 million adults in the UKThe Lancet. Diabetes & Endocrinology, 6
C. Patsa, K. Toutouzas, E. Tsiamis, C. Tsioufis, A. Spanos, A. Karanasos, A. Michelongona, D. Tousoulis, C. Stefanadis (2013)
Impact of metabolic syndrome on clinical outcomes after new generation drug-eluting stent implantation: the 'obesity paradox' phenomenon is still apparent.Nutrition, metabolism, and cardiovascular diseases : NMCD, 23 4
M. Farhangi, M. Vajdi, M. Najafi (2019)
Dietary acid load significantly predicts 10-years survival in patients underwent coronary artery bypass grafting (CABG) surgeryPLoS ONE, 14
E. Barrett-Connor, K. Khaw (1985)
Is hypertension more benign when associated with obesity?Circulation, 72 1
Wentao Bao (2016)
Multiple imputation with multivariate imputation by chained equation (MICE) package.Annals of translational medicine, 4 2
C. Lavie, R. Milani, H. Ventura (2009)
Obesity and cardiovascular disease: risk factor, paradox, and impact of weight loss.Journal of the American College of Cardiology, 53 21
Dr. Jain (2015)
The obesity paradox
C. Hastie, S. Padmanabhan, R. Slack, A. Pell, K. Oldroyd, A. Flapan, K. Jennings, J. Irving, H. Eteiba, A. Dominiczak, J. Pell (2010)
Obesity paradox in a cohort of 4880 consecutive patients undergoing percutaneous coronary intervention.European heart journal, 31 2
Materko Wollner, Benchimol-Barbosa Roberto, Silva Roncally, Nadal Jurandir, Luis Edil (2017)
Accuracy of the WHO’s body mass index cut-off points to measure gender- and age-specific obesity in middle-aged adults living in the city of Rio de Janeiro, BrazilJournal of Public Health Research, 6
Jin Yang, Yuan-jie Li, Qingqing Liu, Li Li, Aozi Feng, Tianyi Wang, Shuai Zheng, Anding Xu, Jun Lyu (2020)
Brief introduction of medical database and data mining technology in big data eraJournal of Evidence-Based Medicine, 13
O. Bosello, M. Donataccio, Massimo Cuzzolaro (2016)
Obesity or obesities? Controversies on the association between body mass index and premature mortalityEating and Weight Disorders - Studies on Anorexia, Bulimia and Obesity, 21
L. Gruberg, N. Weissman, R. Waksman, S. Fuchs, R. Deible, E. Pinnow, Lanja Ahmed, K. Kent, A. Pichard, W. Suddath, L. Satler, J. Lindsay (2002)
The impact of obesity on the short-term and long-term outcomes after percutaneous coronary intervention: the obesity paradox?Journal of the American College of Cardiology, 39 4
A. Gitt, H. Drexel, J. Feely, J. Ferrières, J. González-Juanatey, K. Thomsen, Lawrence Leiter, P. Lundman, P. Silva, T. Pedersen, D. Wood, C. Jünger, P. Dellea, V. Sazonov, F. Chazelle, J. Kastelein (2012)
Persistent lipid abnormalities in statin-treated patients and predictors of LDL-cholesterol goal achievement in clinical practice in Europe and CanadaEuropean Journal of Preventive Cardiology, 19
K. Flegal, B. Graubard, D. Williamson, M. Gail (2007)
Cause-specific excess deaths associated with underweight, overweight, and obesity.JAMA, 298 17
A. Antonopoulos, D. Tousoulis (2017)
The molecular mechanisms of obesity paradox.Cardiovascular research, 113 9
C. Lavie, A. Morshedi-Meibodi, R. Milani (2008)
Impact of cardiac rehabilitation on coronary risk factors, inflammation, and the metabolic syndrome in obese coronary patients.Journal of the cardiometabolic syndrome, 3 3
(2013)
Obesity paradox and the heart
S. Chrysant, G. Chrysant (2013)
New insights into the true nature of the obesity paradox and the lower cardiovascular risk.Journal of the American Society of Hypertension : JASH, 7 1
Shlomit Shachar, G. Williams (2017)
The Obesity Paradox in Cancer—Moving Beyond BMICancer Epidemiology, Biomarkers & Prevention, 26
Surya Artham, C. Lavie, R. Milani, H. Ventura (2008)
The Obesity Paradox: Impact of Obesity on the Prevalence and Prognosis of Cardiovascular DiseasesPostgraduate Medicine, 120
F. Messerli, K. Sundgaard-Riise, E. Reisin, G. Dreslinski, F. Dunn, E. Frohlich (1983)
Disparate cardiovascular effects of obesity and arterial hypertension.The American journal of medicine, 74 5
T. Terada, M. Forhan, C. Norris, Weiyu Qiu, R. Padwal, Arya Sharma, J. Nagendran, Jeffrey Johnson (2017)
Differences in Short‐ and Long‐Term Mortality Associated With BMI Following Coronary RevascularizationJournal of the American Heart Association: Cardiovascular and Cerebrovascular Disease, 6
B. Stähli, C. Gebhard, M. Gick, C. Herman, Miroslaw Ferenc, K. Mashayekhi, H. Buettner, A. Toma (2016)
Impact of body mass index on long-term mortality in women and men undergoing percutaneous coronary intervention for chronic total occlusion.International journal of cardiology, 224
Alistair Johnson, T. Pollard, Lu Shen, Li-wei Lehman, M. Feng, M. Ghassemi, Benjamin Moody, Peter Szolovits, L. Celi, R. Mark (2016)
MIMIC-III, a freely accessible critical care databaseScientific Data, 3
W. Ashton, K. Nanchahal, D. Wood (2001)
Body mass index and metabolic risk factors for coronary heart disease in women.European heart journal, 22 1
Kuldeepa Ghoorah, Peter Campbell, A. Kent, A. Maznyczka, V. Kunadian (2016)
Obesity and cardiovascular outcomes: a reviewEuropean Heart Journal: Acute Cardiovascular Care, 5
L. Gruberg, N. Mercado, S. Milo, E. Boersma, C. Disco, G. Es, P. Lemos, Margalit Tzvi, W. Wijns, F. Unger, R. Beyar, P. Serruys (2005)
Impact of body mass index on the outcome of patients with multivessel disease randomized to either coronary artery bypass grafting or stenting in the ARTS trial: The obesity paradox II?The American journal of cardiology, 95 4
E. Benjamin, S. Virani, C. Callaway, A. Chamberlain, Alexander Chang, Susan Cheng, Stephanie Chiuve, M. Cushman, F. Delling, R. Deo, S. Ferranti, J. Ferguson, M. Fornage, C. Gillespie, C. Isasi, M. Jimenez, L. Jordan, S. Judd, D. Lackland, J. Lichtman, L. Lisabeth, Simin Liu, C. Longenecker, P. Lutsey, J. Mackey, D. Matchar, K. Matsushita, M. Mussolino, K. Nasir, M. O’Flaherty, L. Palaniappan, A. Pandey, D. Pandey, M. Reeves, M. Ritchey, C. Rodriguez, Gregory Roth, W. Rosamond, Uchechukwu Sampson, G. Satou, Svati Shah, Nicole Spartano, D. Tirschwell, C. Tsao, Jenifer Voeks, J. Willey, J. Wilkins, Jason Wu, Heather Alger, Sally Wong, P. Muntner (2018)
Heart Disease and Stroke Statistics—2018 Update: A Report From the American Heart AssociationCirculation, 137
Hindawi Journal of Interventional Cardiology Volume 2021, Article ID 3867735, 10 pages https://doi.org/10.1155/2021/3867735 Research Article Obesity Paradox of All-Cause Mortality in 4,133 Patients Treated with Coronary Revascularization 1,2 1,2 1,2 1,3 1 Chengzhuo Li , Didi Han , Fengshuo Xu , Shuai Zheng , Luming Zhang , 4 1,2 1 1,2 Zichen Wang , Rui Yang , Haiyan Yin , and Jun Lyu Intensive Care Unit, e First Affiliated Hospital of Jinan University, Guangzhou 510630, Guangdong Province, China School of Public Health, Xi’an Jiaotong University Health Science Center, Xian 710061, Shaanxi, China School of Public Health, Shaanxi University of Chinese Medicine, Xianyang 712046, Shaanxi, China Department of Public Health, University of California, Irvine 92697, California, USA Correspondence should be addressed to Haiyan Yin; yinhaiyan1867@126.com and Jun Lyu; lyujun2020@jnu.edu.cn Received 5 August 2021; Accepted 2 November 2021; Published 18 November 2021 Academic Editor: Leonardo De Luca Copyright © 2021 Chengzhuo Li et al. ,is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Objectives. ,e purpose of this study was to determine whether there is a dose-response relationship between body mass index (BMI) and all-cause mortality in patients after coronary revascularization. Methods. ,e MIMIC-III database (version 1.4) was used as the sample population. For variables with less than 10% of values missing, we used the mice package of R software for multiple imputations. Cox regression was used to determine the risk factors of all-cause mortality in patients. RCSs were used to observe the relationship between BMI and all-cause mortality. Additional subgroup and sensitivity analyses were also performed to explore whether the conclusion can be applied to specific groups. Results. Both univariate and multivariate Cox models indicated that the mortality risk was lower for overweight patients than for normal-weight patients (P< 0.05). In RCS models, BMI had a U-shaped relationship with all-cause mortality of patients after coronary artery bypass grafting (CABG) (P for non- linearity � 0.0028). ,ere was a weak U-shaped relationship between BMI and all-cause mortality after percutaneous coronary intervention (PCI), but the nonlinear relationship between these two parameters was not significant (P for nonlinearity � 0.1756). Conclusions. ,e obesity paradox does exist in patients treated with CABG and PCI. RCS analysis indicated that there was a U- shaped relationship between BMI and all-cause mortality in patients after CABG. After sex stratification, the relationship between BMI and all-cause mortality in male patients who received PCI was L-shaped, while the nonlinear relationship among females was not significant. Gruberg et al. were the first to mention the obesity 1. Introduction paradox in patients after revascularization [5]. ,at study Coronary artery revascularization is an important method found that patients with a normal BMI were found to have for treating coronary heart disease. Coronary artery bypass higher hospitalization and 1-year mortality rates than those grafting (CABG) and percutaneous coronary intervention who were overweight or obese. ,is indicates that being overweight or obese may protect patients and reduce (PCI) are the two most common operations in coronary revascularization. ,e body mass index (BMI) is a prog- mortality. However, some studies have attributed the obesity nostic factor for all-cause mortality in many diseases, with paradox to factors other than BMI [6, 7]. the prognosis being better in people with a normal BMI than ,e relationship between BMI and all-cause mortality in in those who are overweight or obese [1, 2]. However, studies patients after coronary revascularization remains unclear. have indicated that some overweight or even obese patients Most studies have included BMI as a continuous or cate- have a better prognosis than those with a normal BMI, which gorical variable when performing logistic or Cox regression is referred to as the obesity paradox [3, 4]. analysis, which does illustrate changes in BMI effectively 2 Journal of Interventional Cardiology 2.2. Statistical Analysis. All data were extracted using the [8–10]. However, basic research is still needed to explore this relationship. Structured Query Language (SQL). Since BMI was the main study indicator, all missing values were deleted, and other ,e purpose of this study was to determine whether there is a dose-response relationship between BMI and all- possible confounding factors with >10% of values missing cause mortality in coronary heart disease patients under- were also deleted. We used the mice package of R software going coronary artery revascularization, thereby determin- for multiple imputation for variables with <10% of values ing whether there is an obesity paradox in the prognosis of missing [13]. ,is package uses relevant random samples patients who underwent coronary artery revascularization. based on the distribution of predicted values to replace the We hypothesized that it is a nonlinear relationship and used missing values with estimated values. We constructed a restricted cubic splines (RCSs) to determine the dose-re- dataset by merging five imputation datasets. sponse relationship. ,e World Health Organization (WHO) standards were used to divide BMI into underweight, normal weight, overweight, and obese categories. Age was divided into three 2. Methods categories: youth, middle age, and elderly. All categorical 2.1. Study Design and Population. ,e MIMIC-III database variables were expressed as numbers and percentages, and the chi-squared and Fisher’s exact tests were used to de- contains data on 53,423 adult patients (16 years or older) admitted to an intensive care unit between 2001 and 2012. termine the differences between the two groups. All con- ,e MIMIC-III database (version 1.4) was sampled, and tinuous variables were expressed as medians and the project was approved by the Institutional Review interquartile-range values, and the Mann–Whitney U test Boards of the Beth Israel Deaconess Medical Center was used to identify differences between the two groups. (Boston, Massachusetts) and the Massachusetts Institute ,e Kaplan–Meier curve and log-rank test were used to of Technology (Cambridge, Massachusetts). Consent was evaluate whether there are differences in survival rates be- tween different operations. Cox regression was used to not required from individual patients since all of the protected health information of the project has been determine the risk factors of patient all-cause mortality. We constructed univariate models, a demographic adjustment deidentified [11, 12]. We completed recognized courses for protecting human research participants, including the model (model I), and a multivariate adjustment model requirements of the HIPAA (Health Insurance Portability (model II). A trend test was used to determine linear trends and Accountability Act), and signed a data usage agree- between BMI and all-cause mortality. ment. ,e Institutional Review Boards of the Beth Israel RCSs were used to assess the nonlinear relationship Deaconess Medical Center and Massachusetts Institute of between BMI and all-cause mortality. We divided the dis- Technology have approved the use of the MIMIC-III tribution of BMI into quartiles as the knots of the nonlinear database by any researcher who meets the data user re- models. We analyzed the nonlinear relationship between quirements, and the requirement for patient’s informed BMI and patient all-cause mortality and determined the P value of the nonlinear test. consent is waived. We used the ninth edition of the International Classi- We also performed a subgroup analysis of sex and age to determine whether a relationship existed within a fication of Diseases (Clinical Modification) codes to identify and analyze all patients in the MIMIC-III database whose specific subgroup. We performed the following sensitivity primary operation was revascularization (including CABG analyses to determine the stability of our results: (1) we and PCI). BMI was calculated using the heights and weights removed all patients who died within 30 days of their recorded in the database. As the main study indicator, all operation and performed the RCS analysis again, (2) missing values for BMI were deleted. ,e MIMIC-III da- abnormal BMI values (more than three times the standard tabase is connected to the social security database to record deviation) were deleted before the analysis, and (3) we deleted the missing values from the original data before the follow-up times and outcomes of patients. Our study outcome was all-cause mortality as registered by the social performing multiple imputations and reanalyzed the data to explore the difference resulting from multiple security bureau of the patient. ,e follow-up times were reported in days. imputations. All statistical analyses were conducted using Navicat We also extracted patient demographic, laboratory, and vital-sign indicators from the database. ,ese factors were Premium and R software (version 3.6.2, ,e R Foundation possible confounding factors in the relationship between for Statistical Computing, Vienna, Austria). All cited P BMI and all-cause mortality. All indicators were extracted values were two-sided, and P< 0.05 was considered statis- from the diagnoses_icd, admissions, patients, icustays, tically significant. labevents, and procedures_icd parameters in the database. ,e flowchart for data inclusion and exclusion is displayed in 2.3. Patient and Public Involvement. ,is research was done Figure 1. without patient involvement. Patients were not invited to ,e purpose of our research was to determine the dose- comment on the study design and were not consulted to response relationship between BMI and all-cause mortality develop patient-relevant outcomes or interpret the results. in patients after revascularization and to determine the Patients were not invited to contribute to the writing or existence of the obesity paradox while adjusting for possible editing of this document for readability or accuracy. confounding factors. Journal of Interventional Cardiology 3 MIMIC-III database Inclusion criteria: (i) Primary operation was revascularization (including CABG and PCI) (ii) Admitted to the ICU for the first time. (iii) All indicators are taken from results within 24 hours (iv) Patients older than 18 years old. Coronary revascularization patients under the above criteria (n=4880) Exclusion criteria: (i) BMI unknown (n = 726) (ii) Laboratory and vital-sign indexes unknown (n = 20) (iii) Survival time unknown (n = 1) Included primary cohort (n = 4133) CABG (n = 3593, 86.9%) PCI (n = 540, 13.1%) Figure 1: ,e screening flowchart. from the multivariate Cox analysis adjusted). Table 2 lists the 3. Results hazard ratio (HR) and 95% confidence interval (CI) values 3.1. Population Characteristics. Our final sample comprised for CABG patients according to their BMI group. Both the 4133 patients treated with coronary artery revascularization univariate model and model II showed that the risk of in the MIMIC-III database: CABG patients (86.9%, mortality was lower for overweight patients than for normal- n � 3593) and PCI patients (13.1%, n � 540). ,eir baseline weight patients (P< 0.05). Table 3 lists the HRs and 95% CIs data are listed in Table 1, which indicates that most of the for PCI patients. All three models indicated that the mor- variables differed significantly between the two types of tality risk was lower for overweight PCI patients than for surgery (P< 0.05). Table 1 indicates that there was a very normal-weight patients (P< 0.05). ,e trend tests of the six small number of underweight patients, with overweight models revealed that a significant linear trend was only patients comprising the largest proportion (approximately present for the BMI univariate model of PCI patients 39%). Most patients were male, married, and white. Figure 2 (P � 0.002), which also suggested the presence of a non- displays the survival curves after surgery for both types of linear relationship between BMI and all-cause mortality. patients. ,e figure suggests that there are significant sta- tistical differences between the two, and so the two types of patients were analyzed separately. 3.3. RCS Analyses of Nonlinear Relationships. Supplementary Tables S3 and S4 list the HR and 95% CIs of the BMI quartiles for all-cause mortality. ,e first quartile 3.2. Univariate and Multivariate Cox Regression Analyses. (Q1) was taken as a reference group for comparison with the We first divided BMI into four categories based on the WHO other groups to obtain the corresponding HRs. In CABG standards and then performed univariate and multivariate patients, compared with the Q1 of BMI, the HR for Q2 of Cox regression analyses with normal weight as the reference. BMI was 0.722 (95% CI � 0.586–0.888) for the univariate model, 0.759 (95% CI � 0.616–0.936) for model I, and 0.740 After adjusting all of the variables, the Cox regression model showed that length of stay, Elixhauser Comorbidity Index (95% CI � 0.599–0.913) for model II (all P< 0.05). Unlike (ECI), urine output, pH, potassium, prothrombin time, red CABG patients, the Q2 group of PCI patients did not show blood cell distribution width (RDW), age, marital status, and significant differences, whereas the Q3 group did show BMI were the significant prognostic factors for CAGB pa- significant differences. Compared with the Q1 of BMI, the tients, while ventilator support, ECI, urine output, heart rate, HR for Q3 of BMI was 0.506 (95% CI � 0.332–0.773) for the temperature, hematocrit, RDW, sex, and BMI were the univariate model, 0.572 (95% CI � 0.370–0.884) for model I, significant prognostic factors for PCI patients. ,e Cox and 0.601 (95% CI � 0.392–0.924) for model II (all P< 0.05). regression results are presented in detail in Supplementary We constructed RCS models to further analyze the re- Tables S1 and S2. lationship between BMI and patient mortality. Figure 3 Tables 2 and 3 present the results for model I (with the shows the dose-response curves of BMI and all-cause mortality after adjusting for the factors that were significant demographic characteristics of age, sex, race, and marital status adjusted) and model II (with the risk factors derived in the Cox analysis. ,e dose-response analysis revealed a 4 Journal of Interventional Cardiology Table 1: Baseline characteristics. Variables Total (n � 4133) CABG (n � 3593) PCI (n � 540) P value Age (n) (%) <0.001 Youth 1847 (44.69) 1601 (44.56) 246 (45.56) Middle aged 1733 (41.93) 1567 (43.61) 166 (30.74) ,e elder 553 (13.38) 425 (11.83) 128 (23.70) Sex (n) (%) <0.001 Male 3113 (75.32) 2751 (76.57) 362 (67.04) Female 1020 (24.68) 842 (23.43) 178 (32.96) Race (n) (%) <0.001 White 2840 (68.72) 2446 (68.08) 394 (72.96) Black 131 (3.17) 100 (2.78) 31 (5.74) Asian 83 (2.01) 76 (2.12) 7 (1.30) Hispanic or Latino 101 (2.44) 92 (2.56) 9 (1.67) Others 978 (23.66) 879 (24.46) 99 (18.33) Insurance (n) (%) 0.014 Government 83 (2.01) 70 (1.95) 13 (2.41) Medicaid 183 (4.43) 163 (4.54) 20 (3.70) Medicare 2241 (54.22) 1915 (53.30) 326 (60.37) Private 1610 (38.95) 1432 (39.86) 178 (32.96) Self-pay 16 (0.39) 13 (0.36) 3 (0.56) Marital (n) (%) <0.001 Married 2674 (64.70) 2368 (65.91) 306 (56.67) Unmarried 559 (13.53) 461 (12.83) 98 (18.15) DSW 729 (17.64) 609 (16.95) 120 (22.22) Others 171 (4.14) 155 (4.31) 16 (2.96) BMI (n) (%) 0.001 I: normal 953 (23.06) 817 (22.74) 136 (25.19) II: underweight 46 (1.11) 32 (0.89) 14 (2.59) III: overweight 1618 (39.15) 1406 (39.13) 212 (39.26) IV: obesity 1516 (36.68) 1338 (37.24) 178 (32.96) Length of stay (day), median (IQR) 2.07 (1.20, 3.30) 2.10 (1.20, 3.29) 1.90 (1.15, 3.67) 0.165 ECI, median (IQR) 6.00 (−2.00, 13.00) 6.00 (−2.00, 11.00) 9.00 (1.00, 19.00) <0.001 Urine output (ml), median (IQR) 2088.00 (1510.00, 2858.00) 2110.00 (1543.00, 2858.00) 1912.50 (1268.75, 2833.75) <0.001 Heartrate (min-1), median (IQR) 84.00 (77.73, 90.48) 84.77 (78.97, 90.91) 75.31 (67.71, 85.28) <0.001 SBP (mmHg), median (IQR) 112.08 (106.53, 119.53) 111.77 (106.58, 118.81) 115.06 (106.03, 125.23) <0.001 DBP (mmHg), median (IQR) 57.15 (53.08, 61.83) 56.71 (52.83, 60.93) 62.48 (55.84, 69.78) <0.001 Mean BP (mmHg), median (IQR) 74.82 (71.07, 79.33) 74.59 (71.07, 78.69) 77.56 (71.01, 83.59) <0.001 Resprate (min-1), median (IQR) 16.83 (15.30, 18.78) 16.68 (15.17, 18.54) 18.17 (16.31, 20.25) <0.001 Temperature ( C), median (IQR) 36.82 (36.52, 37.14) 36.83 (36.53, 37.15) 36.69 (36.43, 37.03) <0.001 SpO (%), median (IQR) 98.13 (97.13, 98.96) 98.23 (97.30, 99.00) 97.14 (95.97, 98.29) <0.001 Free calcium (mmol/L), median (IQR) 1.15 (1.12, 1.19) 1.15 (1.12, 1.19) 1.13 (1.09, 1.18) <0.001 pCO (mmHg), median (IQR) 41.00 (37.00, 45.00) 41.00 (38.00, 45.00) 41.50 (37.00, 46.00) 0.232 pH, median (IQR) 7.40 (7.37, 7.44) 7.41 (7.38, 7.44) 7.38 (7.33, 7.42) <0.001 pO (mmHg), median (IQR) 380.00 (269.00, 433.00) 390.00 (311.00, 437.00) 155.00 (88.00, 338.25) <0.001 Anion gap (mEq/L), median (IQR) 13.00 (11.00, 15.00) 13.00 (11.00, 15.00) 15.00 (13.00, 17.25) <0.001 Bicarbonate (mEq/L), median (IQR) 25.00 (23.00, 27.00) 25.00 (23.00, 27.00) 24.00 (21.75, 26.00) <0.001 Chloride (mEq/L), median (IQR) 105.00 (102.00, 108.00) 105.00 (102.00, 109.00) 103.00 (100.00, 106.00) <0.001 Creatinine (mEq/L), median (IQR) 0.90 (0.80, 1.20) 0.90 (0.80, 1.10) 1.10 (0.80, 1.40) <0.001 Glucose (mg/dL), median (IQR) 125.00 (105.00, 157.00) 123.00 (104.00, 153.00) 138.00 (112.00, 186.25) <0.001 Magnesium (mg/dL), median (IQR) 2.00 (1.80, 2.20) 2.00 (1.80, 2.20) 1.90 (1.80, 2.10) <0.001 Potassium (mg/dL), median (IQR) 4.20 (3.90, 4.50) 4.20 (3.90, 4.50) 4.15 (3.80, 4.60) 0.679 Sodium (mg/dL), median (IQR) 139.00 (137.00, 141.00) 139.00 (137.00, 141.00) 138.00 (136.00, 140.00) <0.001 Urea nitrogen (mg/dL), median (IQR) 17.00 (14.00, 23.00) 17.00 (14.00, 22.00) 19.00 (14.00, 28.00) <0.001 Hematocrit (%), median (IQR) 34.90 (30.10, 39.20) 34.60 (29.70, 38.90) 37.00 (32.40, 40.90) <0.001 Hemoglobin (g/dL), median (IQR) 12.00 (10.19, 13.60) 11.90 (10.19, 13.50) 12.60 (10.90, 14.10) <0.001 INR, median (IQR) 1.20 (1.10, 1.40) 1.20 (1.10, 1.40) 1.20 (1.10, 1.30) <0.001 Platelet count (K/uL), median (IQR) 206.00 (162.00, 257.00) 201.00 (159.00, 250.00) 242.00 (195.00, 297.00) <0.001 PT (s), median (IQR) 13.50 (12.70, 14.80) 13.50 (12.70, 14.80) 13.30 (12.40, 14.60) <0.001 PTT (s), median (IQR) 30.50 (26.70, 39.00) 30.40 (26.80, 37.90) 30.65 (25.50, 57.50) 0.294 RDW (%), median (IQR) 13.60 (13.00, 14.20) 13.50 (13.00, 14.20) 13.90 (13.30, 15.03) <0.001 Red blood cells (m/uL), median (IQR) 3.93 (3.36, 4.45) 3.89 (3.32, 4.42) 4.18 (3.59, 4.65) <0.001 Journal of Interventional Cardiology 5 Table 1: Continued. Variables Total (n � 4133) CABG (n � 3593) PCI (n � 540) P value White blood cells (k/uL), median (IQR) 9.10 (7.00, 12.20) 8.90 (6.90, 12.00) 10.25 (8.30, 13.20) <0.001 Ventilator (n) (%) <0.001 No 738 (17.86) 297 (8.27) 441 (81.67) Yes 3395 (82.14) 3296 (91.73) 99 (18.33) CABG, coronary artery bypass grafting; PCI, percutaneous coronary intervention; IQR, interquartile-range; DSW divorced, separated, or widowed; BMI, body mass index; ECI, Elixhauser comorbidity index; SBP, systolic blood pressure; DBP, diastolic blood pressure; MBP, mean blood pressure; INR, in- ternational normalized ratio; PT, prothrombin time; PTT, partial thromboplastin time; RDW, red cell distribution width. 1.00 0.75 0.50 0.25 p < 0.0001 0.00 0 1000 2000 3000 4000 Time Number at risk Operation=CABG 3593 2209 228 72 2 Operation=PCI 540 177 9 0 0 0 1000 2000 3000 4000 Time Strata Operation Operation=CABG CABG Operation=PCI PCI Figure 2: Kaplan–Meier survival curves for all-cause mortality. U-shaped curve between BMI and the risk of all-cause or age. However, Supplementary Table S6 showed that mortality in CABG patients (P for nonlinearity � 0.0028) significant interaction was found in the PCI group for (Figure 3(a)). In contrast, there was a weak U-shaped re- stratification according to sex (P for interaction � 0.006). lationship between BMI and all-cause mortality after PCI, ,is suggests that the impact of BMI on the mortality of PCI but no significant nonlinear relationship between these two patients is affected by sex. variables (P for nonlinearity � 0.1756) (Figure 3(b)). We, therefore, constructed RCS models to perform a sex- stratified analysis. As shown in Figure 4, the dose-response relationship appeared as U-shaped curves for all-cause 3.4. Subgroup Analyses. After adjusting for the corre- mortality of BMI and CABG. However, the nonlinear trend sponding confounding factors, the subgroup analysis of BMI was significant for males (P for nonlinearity � 0.0163) but and all-cause mortality of patients after CABG (Supple- not for females (P for nonlinearity � 0.1367). ,e sex-related mentary Table S5) showed that there was no significant difference appeared to be greater in PCI patients. ,ere was interaction effect of sex (P for interaction � 0.330) or age (P an approximate L-shaped relationship between BMI and all- for interaction � 0.883). ,is indicates that the relationship cause mortality in male patients (P for nonlinearity � 0.0085) between BMI and CABG did not differ significantly with sex but not in female patients (P for nonlinearity � 0.8574). Strata Survival probability 6 Journal of Interventional Cardiology Table 2: Cox regression analyses of the relationship between BMI and all-cause mortality for patients after CABG. Univariate Model I Model II HR (95%CI) HR (95%CI) HR (95%CI) BMI P value P value P value Normal Reference Reference Reference Underweight 2.8003 1.560–5.028 <0.001 3.38 1.872–6.103 <0.001 2.576 1.424–4.660 0.002 Overweight 0.7787 0.646–0.939 0.009 0.875 0.724–1.057 0.166 0.825 0.682–0.998 0.048 Obesity 0.8758 0.731–1.050 0.151 1.09 0.904–1.314 0.366 0.908 0.751–1.097 0.318 P for trend 0.178 0.275 0.294 Model I: adjust with age, sex, race, and marital status. Model II: adjust with age, marital status, length of stay, ECI, urine output, pH, potassium, PT, and RDW. Table 3: Cox regression analyses of the relationship between BMI and all-cause mortality for patients after PCI. Univariate Model I Model II HR (95% CI) HR (95% CI) HR (95% CI) BMI P value P value P value Normal Reference Reference Reference Underweight 0.911 0.415–2.000 0.817 1.027 0.446–2.365 0.950 0.390 0.166–0.916 0.031 Overweight 0.613 0.432–0.871 0.006 0.672 0.469–0.963 0.030 0.675 0.468–0.972 0.035 Obesity 0.591 0.404–0.865 0.007 0.742 0.498–1.108 0.144 0.680 0.458–1.009 0.056 P for trend 0.002 0.056 0.054 Model I: adjust with age, sex, race, and marital status. Model II: adjust with sex, ECI, urine output, heart rate, temperature, hematocrit, RDW, and ventilator. RCS RCS 2.0 P for non-linearity = 0.0028 P for non-linearity = 0.1756 2.0 1.5 1.5 1.0 1.0 0.5 20 30 40 20 30 40 50 BMI BMI (a) (b) Figure 3: Dose-response curves for BMI and all-cause mortality of CABG patients (a) and PCI patients (b). 3.5. Sensitivity Analysis. After excluding patients who died 4. Discussion within 30 days after the operation or had abnormal BMI values, Higher BMI is usually a risk factor for adverse outcomes of the RCS analysis showed that the nonlinear relationship be- cardiovascular disease and various complications [14]. tween BMI and all-cause mortality remained consistent with However, many studies have found that overweight or obese the previous results. ,at is, there was a significant U-shaped people have a survival advantage compared to people who nonlinear relationship in CABG patients, but no significant are overly thin or have a BMI within the normal range, nonlinear relationship in PCI patients. However, analyzing the which is called the obesity paradox [15, 16]. ,ere have been complete data before performing multiple imputations some reports on the obesity paradox in patients after re- revealed a U-shaped relationship between BMI and all-cause vascularization. However, most of these studies simply di- mortality for both CABG and PCI patients (P for nonlinearity vided BMI into four categories according to international <0.05) (Figure 5). It should be noted that there was no sig- standards and segmented the relationship between BMI and nificant nonlinear relationship for PCI patients after per- outcome [17, 18]. In contrast, the present study used RCSs to forming multiple imputation. HR (95%CI) HR (95%CI) Journal of Interventional Cardiology 7 RCS RCS P for non-linearity = 0.1367 P for non-linearity = 0.0163 2.0 2.0 1.5 1.6 1.2 1.0 0.8 20 30 40 20 30 40 BMI BMI (a) (b) RCS RCS P for non-linearity = 0.8574 3 P for non-linearity = 0.0085 20 25 30 35 40 20 25 30 35 40 BMI BMI (c) (d) Figure 4: Dose-response curves for BMI and all-cause mortality of CABG patients by sex group with male (a) and female (b); dose-response curves for BMI and all-cause mortality of PCI patients by sex group with male (c) and female (d). explore the nonlinear relationship between BMI and all- relationship between the BMI and all-cause mortality for cause mortality after coronary revascularization surgery and, CABG patients. For PCI patients, there was a weak U-shaped thereby, expressed the data using continuous and smooth relationship between these variables, but no significant nonlinear relationship between them after performing graphs, which is more intuitive for explaining the overall trend for ORs of BMI than using the traditional segmented multivariate adjustment. analysis. It is especially noteworthy that our subgroup analysis of ,e present results show that the survival rate is higher PCI patients revealed a significant interaction between sex for CABG patients than for PCI patients. In univariate and and BMI, which means that sex may affect the relationship multivariate Cox regression analyses, BMI was a prognostic between BMI and mortality. ,ere was no significant dif- factor for all-cause mortality in patients regardless of ference in the Cox regression results for the relationship whether they received CABG or PCI. In the RCS analysis, the between BMI and all-cause mortality among females. models before and after adjustment all showed a U-shaped Correspondingly, in the RCS analysis, we only observed HR (95%CI) HR (95%CI) HR (95%CI) HR (95%CI) 8 Journal of Interventional Cardiology RCS RCS RCS P for non-linearity = 0.0049 P for non-linearity = 0.0035 2.0 2.00 2.00 P for non-linearity = 0.0077 1.75 1.75 1.5 1.50 1.50 1.25 1.25 1.0 1.00 1.00 0.75 0.75 20 30 40 50 20 30 40 20 30 40 50 BMI BMI BMI (a) (b) (c) RCS RCS RCS 2.5 P for non-linearity = 0.2051 2.5 P for non-linearity = 0.6119 4 P for non-linearity = 0.0191 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 20 30 40 20 25 30 35 40 20 25 30 35 40 BMI BMI BMI (d) (e) (f) Figure 5: Sensitivity analysis of (1) excluding patients who died within 30 days after the CABG (a) and PCI (d); (2) excluding patients who had abnormal BMI values of CABG patients (b) and PCI patients (e); (3) the complete data before performing multiple imputation of CABG patients (c) and PCI patients (f). significant nonlinear relationships in males, which indicates when analyzing the full sample, which further confirms the that the U-shaped relationship may only apply to males and present research conclusions. ,e present research has yielded further evidence for the that further investigations are needed into the relationship for females. existence of the obesity paradox. However, unlike previous studies [21–23], we used RCSs to obtain HR values, and Previous studies have found BMI to be a poor dis- criminator of females with different risks of coronary heart hence, more-precise risk ranges. In CABG patients, taking a disease. Weight and BMI are now considered less important BMI of 25 kg/m as the reference value, patients with a BMI than previously thought [19]. Materko et al. showed that the within the range of 25–32 kg/m had a lower HR, which existing BMI classification standards are not suitable for corresponds to a higher survival rate. ,e reasons for this females, [20] which might explain why the relationship obesity paradox remain to be elucidated, but some possible between BMI and all-cause mortality in females was not explanations are given below. One hypothesis is that obese significant in the present study. patients receive more aggressive drug treatment than nor- In addition, after removing all missing data without mal-weight patients, resulting in better control of their performing multiple imputations in the sensitivity analysis, various indicators [24, 25]. In addition, it has been reported the nonlinear relationship became significant in PCI pa- that compared with normal-weight patients, obese hyper- tients. ,is result might have been due to there being too tensives have a higher cardiac output, enlarged blood vol- many missing values for PCI patients since only 97 patients ume, and lower systemic vascular resistance. ,is means remained after removing the missing values. Compared with that, for a given level of arterial pressure, obese patients will the 3345 CABG patients, the much smaller sample for PCI have lower total peripheral resistance than lean patients, patients might have introduced bias into the study. ,ere- which could improve the survival rate of obese patients fore, the relationship between BMI and all-cause mortality in [26, 27]. It has also been shown that BMI is a poor indicator PCI patients needs to be investigated further with large for categorizing obesity. Indicators such as waist circum- samples in prospective cohorts. ,e results of the other ference and waist-to-height ratio may better reflect the true sensitivity analysis are consistent with the results obtained degree of obesity. [28, 29]. HR (95%CI) HR (95%CI) HR (95%CI) HR (95%CI) HR (95%CI) HR (95%CI) Journal of Interventional Cardiology 9 ,is study was subject to some limitations. Firstly, the Authors’ Contributions MIMIC-III database is from a single center, and so the Chengzhuo Li and Didi Han contributed equally to this present research results need to be verified in a multicenter work. population. Secondly, although we used Cox regression for analysis, cross-sectional studies cannot be used to determine causality, and so future prospective cohort studies are Acknowledgments needed. ,irdly, the large number of missing values for PCI ,is study was supported by ,e National Social Science patients may have made the results unstable. Larger samples Foundation of China (grant/award no. 16BGL183). are needed to confirm the conclusions and sex differences for PCI patients. Supplementary Materials 5. Conclusion Table S1: selected variables by multivariate Cox regression ,e analysis performed in this study of the MIMIC-III analysis for CAGB patients. Table S2: selected variables by database revealed that the obesity paradox does exist in multivariate Cox regression analysis for PCI patients. Table patients treated with CABG and PCI. ,e RCS analysis S3: Cox regression analyses of the relationship between BMI showed that BMI has a U-shaped relationship with all-cause quartiles and all-cause mortality for patients after CABG. mortality in CABG patients, while this relationship is not Table S4: Cox regression analyses of the relationship between significant in PCI patients. However, sex stratification BMI quartiles and all-cause mortality for patients after PCI. revealed an L-shaped relationship between BMI and all- Tables S5: subgroup analysis of the associations between cause mortality in male PCI patients, while there was no BMI and all-cause mortality for patients after CABG. Tables significant nonlinear relationship in females. Multicenter S6: subgroup analysis of the associations between BMI and and large-sample prospective studies are needed to further all-cause mortality of patients after PCI. Supplementary explore sex differences in the obesity paradox. material related to this article can be found in Supple- mentary Tables S1-S6. (Supplementary Materials) Abbreviations References BMI: Body mass index CABG: Coronary artery bypass grafting [1] K. Bhaskaran, I. dos-Santos-Silva, D. A. Leon, I. J. Douglas, PCI: Percutaneous coronary intervention and L. Smeeth, “Association of BMI with overall and cause- RCS: Restricted cubic splines specific mortality: a population-based cohort study of 3·6 million adults in the UK,” e Lancet Diabetes & Endocri- HIPAA: Health Insurance Portability and Accountability nology, vol. 6, no. 12, pp. 944–953, 2018. Act [2] C. H. Kroenke, W. Y. Chen, B. Rosner, and M. D. Holmes, SQL: Structured Query Language “Weight, weight gain, and survival after breast cancer diag- WHO: World Health Organization nosis,” Journal of Clinical Oncology, vol. 23, no. 7, ECI: Elixhauser Comorbidity Index pp. 1370–1378, 2005. RDW: Red blood cell distribution width [3] K. M. Flegal, B. I. Graubard, D. F. Williamson, and M. H. Gail, HR: Hazard ratio “Cause-specific excess deaths associated with underweight, CI: Confidence interval. overweight, and obesity,” Journal of the American Medical Association, vol. 298, no. 17, pp. 2028–2037, 2007. Data Availability [4] J. P. Curtis, J. G. Selter, Y. Wang et al., “,e obesity paradox,” Archives of Internal Medicine, vol. 165, no. 1, pp. 55–61, 2005. Data were fully available at https://physionet.org/content/ [5] L. Gruberg, N. J. Weissman, R. Waksman et al., “,e impact of obesity on the short-term andlong-term outcomes after mimiciii-demo/1.4/. percutaneous coronary intervention: the obesity paradox?” Journal of the American College of Cardiology, vol. 39, no. 4, Ethical Approval pp. 578–584, 2002. [6] S. Strulov Shachar and G. R. Williams, “,e obesity paradox in ,e MIMIC-III database has received ethical approval from cancer-moving beyond bmi,” Cancer Epidemiology Bio- the Institutional Review Boards of both Beth Israel Dea- markers & Prevention, vol. 26, no. 1, pp. 13–16, 2017. coness Medical Center (Boston, MA, USA) and the Mas- [7] O. Bosello, M. P. Donataccio, and M. Cuzzolaro, “Obesity or sachusetts Institute of Technology (Cambridge, MA, USA). obesities? Controversies on the association between body mass index and premature mortality,” Eating and Weight Consent Disorders-Studies on Anorexia, Bulimia and Obesity, vol. 21, no. 2, pp. 165–174, 2016. Informed consent was unnecessary because the MIMIC-III [8] T. Terada, M. Forhan, C. M. Norris et al., “Differences in data are anonymous and publicly available. short- and long-term mortality associated with BMI following coronary revascularization,” Journal of the American Heart Association, vol. 6, no. 4, 2017. Conflicts of Interest [9] M. Abbasalizad Farhangi, M. Vajdi, and M. Najafi, “Dietary ,e authors declare that they have no conflicts of interests. acid load significantly predicts 10-years survival in patients 10 Journal of Interventional Cardiology underwent coronary artery bypass grafting (cabg) surgery,” cholesterol goal achievement in clinical practice in Europe and Canada,” European Journal of Preventive Cardiology, PLoS One, vol. 14, no. 10, Article ID e0223830, 2019. [10] B. E. Stahli, ¨ C. Gebhard, M. Gick et al., “Impact of body mass vol. 19, no. 2, pp. 221–230, 2011. [26] E. Barrett-Connor and K. T. Khaw, “Is hypertension more index on long-term mortality in women and men undergoing percutaneous coronary intervention for chronic total occlu- benign when associated with obesity?” Circulation, vol. 72, no. 1, pp. 53–60, 1985. sion,” International Journal of Cardiology, vol. 224, pp. 305– [27] F. H. Messerli, K. Sundgaard-Riise, E. Reisin, G. Dreslinski, 309, 2016. F. G. Dunn, and E. Frohlich, “Disparate cardiovascular effects [11] J. Yang, Y. Li, Q. Liu et al., “Brief introduction of medical of obesity and arterial hypertension,” e American Journal of database and data mining technology in big data era,” Journal Medicine, vol. 74, no. 5, pp. 808–812, 1983. of Evidence-Based Medicine, vol. 13, no. 1, pp. 57–69, 2020. [28] S. G. Chrysant and G. S. Chrysant, “New insights into the true [12] A. E. W. Johnson, T. J. Pollard, L. Shen et al., “Mimic-iii, a nature of the obesity paradox and the lower cardiovascular freely accessible critical care database,” Scientific Data, vol. 3, risk,” Journal of the American Society of Hypertension, vol. 7, no. 1, p. 160035, 2016. no. 1, pp. 85–94, 2013. [13] Z. Zhang, “Multiple imputation with multivariate imputation [29] A. De Schutter, C. J. Lavie, D. A. Patel, and R. V. Milani, by chained equation (mice) package,” Annals of Translational “Obesity paradox and the heart,” Current Opinion in Clinical Medicine, vol. 4, no. 2, p. 30, 2016. Nutrition and Metabolic Care, vol. 16, no. 5, pp. 517–524, [14] E. J. Benjamin, S. S. Virani, C. W. Callaway et al., “Heart disease and stroke statistics-2018 update: a report from the American heart association,” Circulation, vol. 137, no. 12, pp. e67–492, 2018. [15] A. S. Antonopoulos and D. Tousoulis, “,e molecular mechanisms of obesity paradox,” Cardiovascular Research, vol. 113, no. 9, pp. 1074–1086, 2017. [16] C. J. Lavie, R. V. Milani, and H. O. Ventura, “Obesity and cardiovascular disease,” Journal of the American College of Cardiology, vol. 53, no. 21, pp. 1925–1932, 2009. [17] L. Gruberg, N. Mercado, S. Milo et al., “Impact of body mass index on the outcome of patients with multivessel disease randomized to either coronary artery bypass grafting or stenting in the arts trial: the obesity paradox ii?” e American Journal of Cardiology, vol. 95, no. 4, pp. 439–444, 2005. [18] C. E. Hastie, S. Padmanabhan, R. Slack et al., “Obesity paradox in a cohort of 4880 consecutive patients undergoing percu- taneous coronary intervention,” European Heart Journal, vol. 31, no. 2, pp. 222–226, 2010. [19] W. Ashton, K. Nanchahal, and D. A. Wood, “Body mass index and metabolic risk factors for coronary heart disease in women,” European Heart Journal, vol. 22, no. 1, pp. 46–55, 2001. [20] W. Materko, P. R. Benchimol-Barbosa, A. R. Silva Carvalho, J. Nadal, and E. Luis Santos, “Accuracy of the who’s body mass index cut-off points to measure gender- and age-specific obesity in middle-aged adults living in the city of rio de janeiro, Brazil,” Journal of Public Health Research, vol. 6, no. 2, p. 904, 2017. [21] C. J. Lavie, A. Morshedi-Meibodi, and R. V. Milani, “Impact of cardiac rehabilitation on coronary risk factors, inflam- mation, and the metabolic syndrome in obese coronary pa- tients,” Journal of the CardioMetabolic Syndrome, vol. 3, no. 3, pp. 136–140, 2008. [22] S. M. Artham, C. J. Lavie, R. V. Milani, and H. O. Ventura, “,e obesity paradox: impact of obesity on the prevalence and prognosis of cardiovascular diseases,” Postgraduate Medicine, vol. 120, no. 2, pp. 34–41, 2008. [23] C. Patsa, K. Toutouzas, E. Tsiamis et al., “Impact of metabolic syndrome on clinical outcomes after new generation drug- eluting stent implantation: the ‘obesity paradox’ phenomenon is still apparent,” Nutrition, Metabolism, and Cardiovascular Diseases, vol. 23, no. 4, pp. 307–313, 2013. [24] K. Ghoorah, P. Campbell, A. Kent, A. Maznyczka, and V. Kunadian, “Obesity and cardiovascular outcomes: a re- view,” European Heart Journal: Acute Cardiovascular Care, vol. 5, no. 1, pp. 77–85, 2016. [25] A. K. Gitt, H. Drexel, J. Feely et al., “Persistent lipid abnor- malities in statin-treated patients and predictors of ldl-
Journal of Interventional Cardiology – Hindawi Publishing Corporation
Published: Nov 18, 2021
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.