A novel and simply calculated nutritional index serves as a useful prognostic indicator in patients with coronary artery disease

International Journal of Cardiology 262 (2018) 92–98

Contents lists available at ScienceDirect

International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard

A novel and simply calculated nutritional index serves as a useful prognostic indicator in patients with coronary artery disease Shinichiro Doi a, Hiroshi Iwata a,⁎, Hideki Wada a, Takehiro Funamizu a, Jun Shitara a, Hirohisa Endo a, Ryo Naito b, Hirokazu Konishi c, Shuta Tsuboi c, Manabu Ogita c, Tomotaka Dohi a, Takatoshi Kasai a, Shinya Okazaki a, Kikuo Isoda a, Katsumi Miyauchi a, Hiroyuki Daida a a b c

Department of Cardiovascular Medicine, Juntendo University, Graduate School of Medicine, 2-1-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan Department of Cardiovascular Medicine, Juntendo University, Urayasu Hospital, Urayasu, Chiba, Japan Department of Cardiovascular Medicine, Juntendo University, Shizuoka Hospital, Izunokuni, Shizuoka, Japan

a r t i c l e

i n f o

Article history: Received 5 November 2017 Received in revised form 13 January 2018 Accepted 9 February 2018 Keywords: Nutritional index Prognosis Coronary artery disease

a b s t r a c t Objective: No nutritional index has been firmly established yet in patients with coronary artery disease (CAD). In this study, we propose a simple to calculate nutritional indicator in patients who underwent percutaneous coronary intervention (PCI) by using parameters routinely measured in CAD and evaluated its prognostic implication. Methods: This study is a retrospective observational analysis of a prospective database. The subjects were consecutive 3567 patients underwent their first PCI between 2000 and 2013 at Juntendo University Hospital in Tokyo. The median of the follow-up period was 6.3 years (range: 0–13.6 years). The novel nutritional index was calculated by the formula; Triglycerides (TG) × Total Cholesterol (TC) × Body Weight (BW) Index (TCBI) = TG × TC × BW / 1000 (TG and TC: mg/dl, and BW: kg). Results: The Spearman non-parametric correlation coefficient between TCBI and the most often used conventional nutritional index, Geriatric Nutritional Risk Index (GNRI), was 0.355, indicating modest correlation. Moreover, Unadjusted Kaplan-Meier analysis showed higher all-cause mortality, cardiovascular mortality, and cancer mortality in patients with low TCBI. Consistently, elevation of TCBI was associated with reduced all-cause (hazard ratio: 0.86, 95%CI: 0.77–0.96, p b 0.001), cardiovascular (0.78, 0.66–0.92, p = 0.003), and cancer mortality (0.76, 0.58– 0.99, p = 0.041) in patients after PCI by multivariate Cox proportional hazard analyses. Conclusion: TCBI, a novel and easy to calculate nutrition index, is a useful prognostic indicator in patients with CAD. © 2018 Elsevier B.V. All rights reserved.

1. Introduction The association between malnutrition and an increased risk of adverse outcomes in patients with various cardiovascular diseases, such as heart failure [1], peripheral artery disease [2,3] and coronary artery disease (CAD) [4,5] has been recognized. While the underlying mechanisms of this association have yet to be determined due to the complex pathophysiology of cardiovascular diseases, frailty may be one of the critical elements in the link between malnutrition and a poor prognosis in cardiovascular disease [6]. Although the simplest and most frequently used indicator of nutritional status in cardiovascular disease in clinical settings is body mass index (BMI), its prognostic implication is still a subject of intense debate in terms of the controversy regarding the “obesity paradox” [7–9]. A number of other questionnaires, scoring systems, and indices for the assessment of nutritional status have been proposed. However, ⁎ Corresponding author at: Department of Cardiovascular Medicine, Juntendo University, Graduate School of Medicine, Hongo 2-1-1, Bunkyo-ku, Tokyo 113-0033, Japan. E-mail address: [email protected] (H. Iwata).

https://doi.org/10.1016/j.ijcard.2018.02.039 0167-5273/© 2018 Elsevier B.V. All rights reserved.

in clinical settings, there are currently no established nutritional indicators that are used in a majority of assessments. Among these indicators, we and others recently reported that the Geriatric Nutritional Risk Index (GNRI) and the Controlling Nutritional Status (CONUT) are useful prognostic indicators for patients with CAD [4,5]. Nonetheless, despite their usefulness as prognostic indicators of cardiovascular disease, GNRI and other indicators are still not commonly calculated, most likely due to the requirement of complex calculations of parameters which are not commonly used in cardiovascular clinical practice. High levels of serum total cholesterol (TC) and triglycerides (TG) and a high body weight (BW)/body mass index (BMI) are known risk factors for the progression of atherosclerosis [10–12]. However, total cholesterol has been paradoxically used in the nutritional indicator CONUT and its reduced level was shown to be associated with higher mortality in patients with stable angina [13], ST segment elevation myocardial infarction [14,15], and chronic heart failure [16]. Moreover, triglycerides have been recently recognized as a candidate for an objective nutritional indicator [17]. The controversy regarding the prognostic effects of low levels of serum TC and TG in assessing cardiovascular risk is similar to that of BW/BMI in the “obesity paradox”.

S. Doi et al. / International Journal of Cardiology 262 (2018) 92–98

In this study, we tested our hypothesis that the simple multiplication of commonly measured objective parameters in patients with CAD; serum TG, serum TC, and body weight, may be a useful indicator of nutritional status, and whether its low level may be associated with adverse outcomes in a patient population with atherosclerosis and CAD. 2. Patients and methods 2.1. Participants and follow-up duration This is a retrospective observational cohort analysis of a prospective database and included consecutive 3567 patients who underwent percutaneous coronary intervention for the first time (index PCI) between January 2000 and December 2013 at Juntendo University Hospital, Tokyo, and whose data regarding TG, TC and BW at index PCI were available. PCI includes any type of coronary artery intervention procedure, such as thrombectomy, balloon angioplasty, and deployment of a bare metal stent (BMS) or drug-eluting stent (DES). Patients who underwent index PCI were followed by chart review when followed at our institution and by sending a prognosis survey questionnaire every 5 years if they were transferred to another institution. If no response was received, follow-up was terminated at the latest time point at which his or her survival at the outpatient clinic or an inpatient ward of our institution was confirmed. The median and range of the follow-up period since the index PCI were 6.3 and 0–13.6 years, respectively. This study was approved by the institutional review board at Juntendo University School of Medicine and written informed consent was obtained from all participants. 2.2. Endpoints The endpoints evaluated were three types of mortality in the follow-up period; allcause mortality, cardiovascular mortality, and cancer mortality. Cardiovascular mortality was defined as death due to myocardial infarction, heart failure, critical arrhythmia, valvular heart disease, an aortic disease, peripheral artery disease, or sudden death for which a non-cardiovascular cause could be excluded. Cancer mortality was defined as death due to any type of malignant disease, including a malignant solid tumor, as well as a hematological malignancy. During follow-up, all-cause mortality occurred in 501 of the 3567 patients (14.0%). Among these 501 patients, the cause of death was cardiovascular in 212 (42.3%) and cancer in 160 (31.9%). The deaths due to malignancies included gastrointestinal cancer in 50 patients (31.3%), lung cancer in 34 (21.2%), hepatobiliary cancer in 27 (16.9%), and prostate cancer in 14 (8.8%). 2.3. Generation of a novel nutritional and prognostic indicator in patients with coronary artery disease To generate a useful and versatile nutritional and prognostic indicator in the field of cardiology, we evaluated candidate nutrition-related parameters that met the following requirements, 1) objectively measurable (i.e., not obtained from a subjective questionnaire) 2) measured in the vast majority of atherosclerotic cardiovascular disease cases in clinical practice, and 3) easy to calculate in a clinical setting. The candidate parameters evaluated were BW, body mass index (BMI), TC, and TG. The association between single, double, and triple combinations of these parameters and the 3 types of mortality after PCI were evaluated in comparison with the conventional nutritional index GNRI and serum albumin. 2.4. Calculation of Geriatric Nutritional Risk Index (GNRI) and a novel nutritional risk index, Triglycerides ∗ Total Cholesterol ∗ Body Weight Index (TCBI) GNRI = 14.89 × serum Alb (g/dL) + 41.7 × (measured body weight (kg)/ideal body weight (kg)) [18]. Ideal body weight was calculated using the Lorentz-formula. Ideal body weight = (height (cm) − 100) − (height (cm) − 150) / 4 for men and (height (cm) − 100) − (height (cm) − 150) / 2 for women [18]. TCBI ¼ serum triglycerides ðTG; mg=dLÞ  serum total cholesterol ðTC; mg=dLÞ  body weight ðBW; kgÞ=1000

2.5. Statistical analysis Continuous valuables are presented as the mean ± standard deviation or median with interquartile range in accordance with the results of the Shapiro-Wilk normality test, and were compared using the non-parametric Mann-Whitney test. Categorical data are shown as numbers and percentages and were compared using the Fisher exact test. Since TCBI was non-normally distributed, the levels of TCBI by quartiles of GNRI, and those of GNRI by quartiles of TCBI were compared by Nonparametric Kruskal-Wallis analysis. Kaplan-Meier curves for evaluation of the time to the three types of mortality were drawn and followed by log-rank test for comparison. Unadjusted univariate Cox proportional hazard analyses for all-cause mortality, cardiovascular mortality, and cancer mortality were performed (Supplementary Table 2). Multivariate Cox proportional hazard analysis using 4 models with variables identified based on univariate unadjusted analysis

93

(Supplementary Table 3) was performed to identify factors associated with the incidence of mortality. Variables used in the models (Models 2–4) of multivariate analysis were selected in accordance with data background demographics and univariate Cox hazard analysis (Table 1 and Supplementary Table 1). All probability values (p-values) were two-tailed and considered as significant if b0.05.

3. Results 3.1. Baseline data of TCBI and its correlation with Geriatric Nutritional Risk Index (GNRI) The median and interquartile range (IQR) of TCBI were 1309.1 and 857.7–2060.7, respectively. While TCBI was not normally distributed, it was lognormally distributed (Supplementary Fig. 1a). To address the utility of TCBI as a nutritional index, the correlation between TCBI and GNRI was evaluated. The value of GNRI increased with an increase in the quartile of TCBI. Nonparametric Kruskal-Wallis analysis showed significant elevation in GNRI with an increase in quartiles in TCBI (p b 0.001). TCBI increased in accordance with the quartile of GNRI (p b 0.001) (Supplementary Fig. 1b). Both TCBI and GNRI in the elderly (70 years old or older) and females were significantly lower (p b 0.001), while there was no significant difference between patients with diabetes or hypertension (Supplementary Fig. 1c). Spearman's non-parametric correlation coefficient between TCBI and GNRI was 0.355, while those between GNRI and TG or TC were 0.196 and 0.216, respectively (Supplementary Fig. 1d). These findings indicated a moderately positive linear correlation of TCBI with GNRI. 3.2. Baseline patient demographics and lipid parameters by quartiles of TCBI Baseline patient demographics according to the quartiles of TCBI are shown in Table 1. Patients in the lower quartiles of TCBI were likely to be older, female, and have less dyslipidemia, acute coronary syndrome (ACS), multivessel disease, low left ventricular ejection fraction (LVEF) with higher plasma BNP, and chronic kidney disease (CKD) with lower eGFR. Regarding medications at PCI, fewer patients received betablockers and statins in the groups with lower TCBI. Target vessel size was smaller in the lower TCBI groups. Lipid parameters are summarized in Supplementary Table 1. Serum LDL, ApoB100 and ApoE were slightly positively correlated with TCBI, while HDL was inversely correlated with TCBI. 3.3. Unadjusted Kaplan-Meier analyses for evaluating cumulative incidences of all-cause, cardiovascular, and cancer mortality in accordance with quartiles of TCBI Unadjusted Kaplan-Meier curves showed a significantly higher cumulative incidence of all-cause death in the lowest (Q1) and second lowest (Q2) quartiles of TCBI compared to the third (Q3) and highest quartiles (Q4). Also, the all-cause mortality rate in Q2 was lower than Q1, while those in Q3 and Q4 were similar (Fig. 1a, upper panel). The cardiovascular mortality rate in Q1 was higher than that in the other three quartiles, and the rate in Q2, which was lower than Q1, was also higher than Q3 and Q4 (Fig. 1a, middle panel). Similar to those for all-cause death, the Kaplan-Meier curves of Q3 and Q4 were similar. Regarding cancer mortality, Q1 and Q2 were significantly higher than Q3 and Q4 (Fig. 1a, lower panel). 3.4. Kaplan-Meier analyses in patient subclasses consistently showed lower survival rates in patients with lower quartiles of TCBI Kaplan-Meier analysis consistently showed higher all-cause mortality rates in patients with the lowest quartile (Q1) or second lowest (Q2) of TCBI compared to the third (Q3) and fourth (highest, Q4) quartiles in various subclasses of patients, such as with preserved and reduced left

94

S. Doi et al. / International Journal of Cardiology 262 (2018) 92–98

Table 1 Patient and lesion demographics in accordance with quartiles of TCBI. Q1 (N = 892)

Q2 (N = 892)

Q3 (N = 891)

Q4 (N = 892)

P-value

Demographics Age Elderly (N70 y.o.) Gender, male Hypertension SBPb, mm Hg Diabetes HbAlc, % Dyslipidemia Current smoking Family history Acute coronary syndrome Multi vessel disease Number of diseased vessel LVEF b 40% LVEF, % BNP, pg/ml Chronic kidney disease Cre, mg/dl eGFR, ml/min/1.73 ∗ 2 hsCRP, mg/dl

71.5 ± 16.6 477,53.5% 649,72.8% 626,70.2% 132.4 ± 25.3 387,43.4% 6.2 ± 1.1 514,57.8% 152,17.0% 237,26.7% 340,38.1% 543,60.9% 1.9 ± 0.8 88,11.6% 59.4 ± 14.3 242.8 ± 417.9 331,37.1% 1.3 ± 1.7 65.7 ± 26.3 0.9 ± 2.3

67.2 ± 10.0 354,39.7% 722,80.9% 653,73.2% 134.1 ± 22.8 401,45.0% 6.3 ± 1.2 600,67.3% 202,22.7% 282,31.8% 241,27.0% 501,56.2% 1.82 ± 0.83 60,7.8% 61.5 ± 12.4 162.5 ± 317.2 266,29.8% 1.2 ± 1.7 68.4 ± 23.8 0.6 ± 1.8

65.3 ± 9.4 292,32.8% 762,85.5% 637,71.5% 134.7 ± 21.8 410,46.0% 6.3 ± 1.2 669,75.2% 215,24.2% 262,29.5% 232,26.0% 513,57.6% 1.82 ± 0.82 54,7.0% 61.8 ± 12.1 122.7 ± 298.6 281,31.5% 1.3 ± 2.1 67.8 ± 24.3 0.6 ± 1.4

60.8 ± 10.2 161,18.0% 797,89.3% 626,70.2% 135.4 ± 22.3 379,42.5% 6.5 ± 1.4 767,86.2% 342,38.5% 238,26.8% 293,32.8% 477,53.5% 1.72 ± 0.78 46,6.1% 61.5 ± 11.8 90.6 ± 189.0 235,26.3% 1.0 ± 1.3 72.0 ± 22.1 0.5 ± 1.7

b0.001 b0.001 b0.001 0.439 0.040 0.441 b0.001 b0.001 b0.001 0.054 b0.001 0.016 0.001 0.001 0.001 b0.001 b0.001 0.006 b0.001 0.001

Medications Beta blockers Calcium channel blockers ACEIs/ARBs Statin

389,44.3% 328,37.4% 453,51.6% 446,50.8%

419,47.8% 369,42.2% 474,54.1% 516,59.0%

433,49.2% 367,41.7% 435,49.4% 515,58.5%

448,52.0% 309,35.9% 439,51.0% 528,61.3%

0.013 0.014 0.263 b0.001

Target lesion of PCIa LMCA LAD RCA LCx Graft Reference distal diameter Total stented length

33,30.3% 279,26.1% 400,24.4% 165,24.6% 15,19.7% 2.9 ± 0.5 22.3 ± 16.1

28,25.7% 248,23.2% 425,25.9% 168,25.0% 23,30.3% 2.9 ± 0.5 23.0 ± 16.6

30,27.5% 280,26.1% 405,24.7% 154,22.9% 22,28.9% 2.9 ± 0.5 23.6 ± 16.1

30,27.5% 280,26.1% 405,24.7% 154,22.9% 22,28.9% 3.0 ± 0.5 22.1 ± 15.2

a b

0.326

b0.001 0.18

LMCA: left main coronary artery, LAD: left anterior ascending artery, RCA: right coronary artery, LCx: left circumflex artery. SBP: systolic blood pressure.

ventricular ejection fraction (LVEF), with single and multivessel disease, with and without chronic kidney disease (CKD), with and without statin use, and with and without acute coronary syndrome (ACS). Similarly, patients in Q2 of TCBI also had higher mortality compared with those in Q3 and 4 in patients with or without low LVEF, multivessel disease, statin use and ACS. On the contrary, there was no significant difference in the all-cause mortality rate between Q3 and Q4 in most of the subclasses. In a subclass with chronic kidney disease, the all-cause mortality rate in the highest quartile (Q4) of TCBI was significantly higher than that in Q3. (Fig. 1b).

3.5. Unadjusted univariate Cox proportional analysis identified factors associated with all-cause, cardiovascular and cancer death after PCI for multivariate analysis Unadjusted univariate Cox hazard proportional analyses were performed to identify risk factors associated with the 3 different types of mortality after PCI. Age, male gender, low BMI, diabetes, family history of CAD, hemodialysis, ACS, low EF, multivessel disease, non-LAD lesion, no beta-blocker use, no statin use, high hs-CRP, low eGFR, and low serum total protein (TP) and albumin were identified as risk factors for all-cause death. Factors associated with cardiovascular death were similar to those for all-cause mortality. For cancer death, the predictors were age, current smoking, low eGFR, low hemoglobin, low serum TP, and albumin. Based on these results, models of multivariate Cox proportional hazard analysis were prepared (Supplementary Table 2).

3.6. Multivariate Cox proportional hazard analysis to evaluate the prognostic implication of TCBI for predicting all-cause, cardiovascular and cancer death after PCI To address the implication of TCBI as an independent prognostic predictor in patients after PCI, we prepared 4 models of multivariate analysis. Model 1 included age, male gender, and Q1 of TCBI. Based on the results of background demographics (Table 1) and univariate analysis for all-cause and cardiovascular death (Supplementary Table 2), Model 2 included diabetes, chronic hemodialysis, ACS, multivessel disease, betablocker use, statin use, and CKD in addition to the three covariates in Model 1. While TCBI was evaluated as a nominal variable in Models 1 and 2, in Model 3, the hazard ratio of one standard deviation (1SD) elevation was calculated in combination with 1SD elevation in LVEF, eGFR, and serum albumin as continuous variables, in addition to age (a continuous variable in all models), male gender, diabetes, ACS, multivessel disease, use of statins, and use of beta blockers. Model 4 was constructed for cancer death, because factors associated with cancer death were considerably different from those for cardiovascular death. We basically included variables in Models 2–4 that were 1) indicated as a significant predictor for allcause and cardiovascular death in univariate analysis, and which were significantly different in background demographics between Q1 and Q2–4. Consequently, Q1 of TCBI was significantly associated with an increased risk of all-cause mortality in multivariate analyses by using Models 1 and 2. One standard deviation elevation of TCBI, as a continuous variable, was consistently and significantly associated with a reduced risk of all-cause mortality (Table 2a-1 and a-2). Similarly, Q1 of TCBI was associated with an increased risk, and its 1SD elevation was

S. Doi et al. / International Journal of Cardiology 262 (2018) 92–98

95

Fig. 1. a. Higher incidence of cumulative all-cause, cardiovascular and cancer mortality in patients with lower TCBI. Unadjusted Kaplan-Meier analyses constantly showed higher all-cause, cardiovascular and cancer mortality in patients with lower quartiles (Q1 and Q2) of TCBI compared to higher quartiles (Q3 and Q4) of TCBI. b. Unadjusted Kaplan-Meier analyses indicating cumulative incidence of all-cause mortality by quartiles of TCBI in subclasses. Higher cumulative incidence of all-cause mortality in patients with lowest quartile (Q1) or second lowest quartile (Q2) of TCBI, compared to those with third and highest quartiles of TCBI (Q3 and Q4), was observed in patient subgroups with and without reduced ejection fraction (EF) (a), multivessel disease (b), chronic kidney disease (c), statin use (d), and acute coronary syndrome (e). *,**, *** Indicate 0.01 ≤ p b 0.05, 0.001 ≤ p b 0.01, p b 0,001 in comparison with Q1, respectively. §, §§, §§§ Indicate 0.01 ≤ p b 0.05, 0.001 ≤ p b 0.01, p b 0,001 in comparison with Q2, respectively.

associated with a decreased risk for cardiovascular mortality (Table 2b1 and b-2). For cancer death, 1SD elevation of TCBI predicted a decreased risk in Models 1 and 4 (Table 2c-2), while Q1 of TCBI did not have a significant association with cancer death in Model 4 (Table 2c-1). 3.7. Prognostic implication of NCBI in comparison with other nutritionrelated parameters Unadjusted univariate and adjusted multivariate Cox proportional hazard analysis by multiple models were used to determine the

prognostic implications of single, double and triple combinations of TG, TC, and BW for all-cause mortality and cardiovascular mortality (Supplementary Table 4). Triple combination of these included TCBI and TG × TC / BW. Consequently, the prognostic significance of TCBI remained constant for both mortalities in nominal and continuous variables. In addition, the hazard ratios of TCBI were compared with those of another TCBI, calculated using body mass index (BMI) (TCBI2), and revealed no significant advantage of TCBI2 as an indicator for all-cause, cardiovascular and cancer mortality as a nominal and a continuous (Supplementary Table 5). Finally, we measured the area under the receiver operating

96

S. Doi et al. / International Journal of Cardiology 262 (2018) 92–98

Table 2 Unadjusted and adjusted hazard ratios of lowest quartile of TCBI (as a nominal variable) and 1 standard deviation (1SD) elevation of TCBI (as a continuous variable) for three kinds of mortalities. a. Hazard ratios for all cause death

Hazard ratio

95% CI

p-Value

1. Lowest quartile of TCBI (as a nominal variable) Univariate, unadjusted 2.16 Multivariate, adjusted by Model 1 1.95 Multivariate, adjusted by Model 2 1.58

(1.80–2.59) (1.62–2.35) (1.29–1.94)

b0.001 b0.001 b0.001

2. 1SD elevation of TCBI (as a continuous variable) Univariate, unadjusted 0.62 Multivariate, adjusted by Model 1 0.65 Multivariate, adjusted by Model 3 0.86

(0.57–0.68) (0.59–0.71) (0.77–0.96)

b0.001 b0.001 0.009

b. Hazard ratios for cardiovascular death

95% CI

p-Value

1. Lowest quartile of TCBI (as a nominal variable) Univariate, unadjusted 2.48 Multivariate, adjusted by Model 1 1.95 Multivariate, adjusted by Model 2 1.75

(1.88–3.25) (1.62–2.35) (1.29–2.37)

b0.001 b0.001 b0.001

2. 1SD elevation of TCBI (as a continuous variable) Univariate, unadjusted 0.58 Multivariate, adjusted by Model 1 0.61 Multivariate, adjusted by Model 3 0.78

(0.51–0.67) (0.53–0.71) (0.66–0.92)

b0.001 b0.001 0.003

c. Hazard ratios for cancer death

Hazard ratio

95% CI

p-Value

1. Lowest quartile of TCBI (as a nominal variable) Unadjusted univariate 1.53 Multivariate, adjusted by Model 1 1.46 Multivariate, adjusted by Model 4 1.00

Hazard ratio

(1.09–2.15) (1.03–2.07) (0.58–1.73)

0.014 0.034 0.989

2. 1SD elevation of TCBI (as a continuous variable) Univariate, unadjusted 0.71 Multivariate, adjusted by Model 1 0.71 Multivariate, adjusted by Model 4 0.76

(0.60–0.83) (0.61–0.84) (0.58–0.99)

0.014 0.034 0.041

Models used in multivariate Cox hazard analysis were summarized in Supplementary Table 3. Model 1 includes age, gender, and lowest quartile (1Q) of log TCBI or 1SD elevation of log TCBI. Model 2 includes age, gender, diabetes, chronic hemodialysis, acute coronary syndromes, multi vessel disease, use of beta blockers, use of statins, chronic kidney disease (stages 3–5) and 1Q of log TCBI. Model 3 includes age, gender, diabetes, acute coronary syndromes, multivessel disease, use of beta blockers, use of statins, LVEF (continuous), eGFR (continuous), serum albumin (continuous) and 1SD elevation of log TCBI (continuous). Model 4 includes age, gender, current smoking and hemoglobin (continuous) and 1SD elevation of log TCBI.

characteristic (ROC) curves of TG, TC, BW, GNRI, serum albumin, and TCBI2 in addition to TCBI, and found no significant inferiority in the prognostic implication of TCBI compared to other parameters including serum albumin and GNRI (Supplementary Table 6 and Supplementary Fig. 2). 3.8. Lowest quartile in TCBI was a predictor for all-cause mortality and cardiovascular mortality in various subgroups To evaluate risk stratification having the lowest quartile of TCBI in various subclasses of patients undergoing PCI, we performed multivariate Cox proportional hazard analyses using a model that included age, male gender, and 1SD elevation of TCBI, but excluding the subgroups of female/male and age 70 years old or greater (in these subgroups, gender or age was excluded from the model, respectively). Age and TCBI were considered as continuous variables. Fig. 2a shows that 1SD elevation in TCBI was continuously associated with a decreased risk of all-cause mortality, such as whether male or female, whether 70 years old or greater or not, patients with or without hypertension, diabetes, ACS, low LVEF, multivessel disease, statin use, and LDL cholesterol of 130 mg/dL or lower. Similarly, it was a significant predictor for a decreased risk of cardiovascular death in subgroups (Fig. 2b), such as males and females, whether 70 years old or older or not, patients with or without diabetes, ACS, preserved LVEF, multivessel disease, statin

use, with hypertension, and 130 mg/dL or lower in LDL cholesterol. The index was not a significant predictor for cardiovascular mortality in patient subgroups without hypertension and low LVEF (b40%). Interestingly, 1SD elevation in TCBI was not associated with the incidence of either all-cause death or cardiovascular death in patients with higher LDL (N130 mg/dL). 3.8.1. Impact of TCBI for all-cause, cardiovascular and cancer mortality was evaluated in a propensity-score matched cohort analysis Among all participants in this study, two cohorts having the lowest quartile (1Q) and second to highest (2–4Q) TCBI and who were exactly matched with respect to age, gender, acute coronary syndrome (ACS) and statin use were identified (1Q vs. 2–4Q: N = 741 in each cohort). As the background demographics show in Supplementary Table 7, age, gender, incidence of ACS and statin use were exactly matched in the 2 groups. Kaplan-Meier analyses showed significantly higher allcause mortality and cardiovascular mortality in the group with 1Q than 2–4Q in TCBI, while no difference in cancer mortality was observed. Multivariate Cox proportional hazard analysis using Models 3 and 4 showed risk reduction of 14%, 24%, and 30% for all-cause, cardiovascular and cancer mortality by 1SD elevation of TCBI, respectively (Supplementary Table 8). These findings in a propensity scorematched cohort were very consistent with the main results. 4. Discussion In this study, we propose a novel indicator of nutritional status that is much simpler to calculate, compared to conventional nutritional indicators, using objective parameters measured in a vast majority of patients with atherosclerotic diseases. Since the unfavorable prognostic impact of a poor nutritional status in healthy and diseased individuals has been clarified, the clinical importance of nutritional status has been increasing. Several scoring systems and indexes have been proposed. The Mini Nutritional Assessment (MNA) is conducted mainly by questionnaire in combination with body mass index (BMI), while the Malnutrition Inflammation Score (MIS) is assessed by questionnaire with total iron binding capacity or transferrin [19]. The CONUT is scored using the serum level of albumin and total lymphocyte count in addition to TC [20], while the Prognostic Nutritional Index (PNI) score is calculated by serum albumin level and total lymphocyte count [21]. GNRI is assessed by serum albumin level, measured BW, and ideal BW as described by the Lorentz formula. These three indexes, CONUT, PNI and GNRI, can be calculated by using objective parameters and were initially developed for evaluating nutritional status in patients with malignant diseases [22]. In cardiovascular diseases, the clinical importance of nutritional status has been historically recognized as “cardiac cachexia” in patients with chronic heart failure [23–25]. Meanwhile, the clinical importance of nutritional status in atherosclerotic disease has been less of a focus, although it has recently gathered attention as a risk factor for adverse outcomes in patients who underwent PCI [5] and in those with peripheral artery disease [2,3]. In this study, to establish a novel nutritional indicator in cardiovascular disease, we employed three cardiologist-friendly objective parameters. The novel nutritional index TCBI was obtained by simply multiplying TC, TG and BW and dividing by 1000. Even though a modest correlation between TCBI and GNRI was demonstrated by Spearman's correlation coefficient, a nonparametric test showed a linear increase in TCBI in accordance with that of GNRI, and vice versa. Consistent with our previous study [4] and other [2,3] previous studies that demonstrated an association between poor nutritional status represented by low GNRI and poor prognosis in patients with atherosclerotic cardiovascular disease, a reduced TCBI was significantly associated with a higher cumulative incidence of all-cause and cardiovascular mortality by unadjusted as well as adjusted multivariate analysis in this study. Subclass analyses consistently revealed a similar prognostic implication for TCBI. Of note, even in patients using statins, which might have a confounding effect on serum

S. Doi et al. / International Journal of Cardiology 262 (2018) 92–98

97

Fig. 2. Hazard ratios of 1 standard deviation (1SD) elevation of TCBI for all-cause and cardiovascular mortality in various subgroups. a. Multivariate analyses for all-cause death demonstrated constant significant risk reduction for all-cause mortality by 1SD elevation of NCBI in most subgroups, with the exception of patients with high LDL cholesterol (N130 mg/dL). b. Similarly, for cardiovascular mortality, 1SD elevation of TCBI was associated with a 30–40% risk reduction, except in patients without hypertension and with low EF (b40%) as well as high LDL (N130 mg/dL). DM: diabetes, HT: hypertension, ACS: acute coronary syndrome, LVEF: left ventricular ejection fraction, MVD: Multivessel disease.

TC levels through a reduction in LDL-cholesterol, low TCBI was an independent predictor for all-cause and cardiovascular death. In a subclass with a high LDL-C level (b130 mg/dL), there was no significant association between TCBI and all-cause and cardiovascular mortality, presumably due to the substantially smaller number of individuals with low TCBI in this subgroup (86 patients with 1Q TCBI out of 896 with LDL N 130 mg/dL (9.6%) vs. 806 out of 2669 without LDL N 130 mg/ dl (30.2%), p b 0.001). Kaplan-Meier analyses consistently showed a significantly higher cumulative incidence of all-cause, cardiovascular and cancer death in the lower two quartiles of TCBI. Multivariate Cox proportional hazard analysis showed that a low level of TCBI was associated with an increased risk of all-cause and cardiovascular mortality. Elevation of the

TCBI was associated with a decreased risk of cancer mortality in addition to all-cause and cardiovascular death. Moreover, as there was a considerable difference in the background demographics of patients with low and high quartiles of TCBI that may be difficult to completely exclude by multivariate analysis, we performed propensity score matching and generated two patient cohorts with the lowest (Q1) and second to highest (Q2–4) quartiles of TCBI completely matched in terms of age, gender, acute coronary syndrome and statin use. Consistent with the main results, a low TCBI induced higher all-cause and cardiovascular mortality, and an elevation in TCBI was associated with reduced mortality risk. These findings in propensity score-matched cohorts confirmed the prognostic implication of TCBI in patients with coronary artery disease.

98

S. Doi et al. / International Journal of Cardiology 262 (2018) 92–98

One of the potential important links mediating an association between poor nutritional status and poor health outcomes may be frailty, which is defined as a clinically recognizable state of increased vulnerability and dysfunction in multiple physiologic systems [6,26]. In particular, a number of studies have previously demonstrated the prognostic impact of frailty on mortality and cardiovascular events in the elderly with chronic heart failure [27] and coronary artery disease [28]. Although this study did not evaluate the degree of frailty of the subjects, a possible inverse link between TCBI and frailty may exist. This study has several limitations. We did not investigate the risk of high TCBI for cardiovascular events, because high values of TG, TC and BW are classical risk factors for the development and progression of atherosclerosis. Thus, another study may need to assess the risk of high TCBI for cardiovascular events, such as myocardial infarction, ischemic stroke and cardiovascular death. Moreover, this study is an observational cohort study evaluating prospective data from a single center in Tokyo. A larger sample size in studies conducted at multiple centers with multiple lesions may be needed to generalize the findings in this study. In addition, the clinical utility of TCBI as a nutritional and prognostic indicator may need to be assessed in another context of cardiovascular diseases, such as heart failure. Additionally, although we showed the prognostic utility of TCBI in a subgroup with statins, the impact of lipid-lowering medications including statins, ezetimibe and PCSK-9 inhibitors on the prognostic impact of TCBI should be addressed in future studies. Finally, as the evaluation of TCBI in this study was only cross sectional, the prognostic effect of temporal changes in TCBI at multiple time points may be able to address the clinical efficacy of nutritional intervention. In conclusion, TCBI, simply calculated by multiplying the serum values of TG, TC and measured BW, is a novel indicator of nutritional status in patients with coronary artery disease who underwent PCI. It was significantly associated with the most often used, but relatively more complicated nutritional index, GNRI. Furthermore, this novel index is also a useful prognostic indicator in these patients to predict all-cause mortality, cardiovascular mortality, and cancer mortality after PCI. Supplementary data to this article can be found online at https://doi. org/10.1016/j.ijcard.2018.02.039. Declaration of competing interest All authors have nothing to declare. Author contributions Shinichiro Doi: Conception and design of the work, Data collection, Data analysis and interpretation. Hiroshi Iwata: Conception and design of the work, Data collection, Data analysis and interpretation, Drafting the article. Hideki Wada: Conception and design of the work, Data collection. Takehiro Funamizu: Data collection. Jun Shitara: Data collection. Hirohisa Endo: Data collection. Ryo Naito: Data collection. Hirokazu Konishi: Data collection. Shuta Tsuboi: Data collection. Manabu Ogita: Data collection. Tomotaka Dohi: Data collection. Takatoshi Kasai: Data collection. Shinya Okazaki: Data collection. Kikuo Isoda: Data collection. Katsumi Miyauchi: Conception and design of the work, Critical revision of the article, Final approval of the version to be published. Hiroyuki Daida: Conception and design of the work, Critical revision of the article, Final approval of the version to be published.

References [1] A. Casas-Vara, F. Santolaria, A. Fernandez-Bereciartua, E. Gonzalez-Reimers, A. Garcia-Ochoa, A. Martinez-Riera, The obesity paradox in elderly patients with heart failure: analysis of nutritional status, Nutrition 28 (2012) 616–622. [2] J.S. Lane, C.P. Magno, K.T. Lane, T. Chan, D.B. Hoyt, S. Greenfield, Nutrition impacts the prevalence of peripheral arterial disease in the United States, J. Vasc. Surg. 48 (2008) 897–904. [3] H. Luo, H. Yang, B. Huang, D. Yuan, J. Zhu, J. Zhao, Geriatric Nutritional Risk Index (GNRI) independently predicts amputation Inchronic Criticallimb Ischemia (CLI), PLoS One 11 (2016), e0152111. [4] A. Kunimura, H. Ishii, T. Uetani, et al., Impact of Geriatric Nutritional Risk Index on cardiovascular outcomes in patients with stable coronary artery disease, J. Cardiol. 69 (2017) 383–388. [5] H. Wada, T. Dohi, K. Miyauchi, et al., Prognostic impact of the geriatric nutritional risk index on long-term outcomes in patients who underwent percutaneous coronary intervention, Am. J. Cardiol. 119 (2017) 1740–1745. [6] Q.L. Xue, The frailty syndrome: definition and natural history, Clin. Geriatr. Med. 27 (2011) 1–15. [7] H.S. Gurm, D.M. Brennan, J. Booth, J.E. Tcheng, A.M. Lincoff, E.J. Topol, Impact of body mass index on outcome after percutaneous coronary intervention (the obesity paradox), Am. J. Cardiol. 90 (2002) 42–45. [8] V. Hainer, I. Aldhoon-Hainerova, Obesity paradox does exist, Diabetes Care 36 (Suppl. 2) (2013) S276–81. [9] S.B. Heymsfield, W.T. Cefalu, Does body mass index adequately convey a patient's mortality risk? JAMA 309 (2013) 87–88. [10] Global Burden of Metabolic Risk Factors for Chronic Diseases C, Y. Lu, K. Hajifathalian, et al., Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1.8 million participants, Lancet 383 (2014) 970–983. [11] Group SSSS, Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S), Lancet 344 (1994) 1383–1389. [12] M. Miller, C.P. Cannon, S.A. Murphy, et al., Impact of triglyceride levels beyond lowdensity lipoprotein cholesterol after acute coronary syndrome in the PROVE IT-TIMI 22 trial, J. Am. Coll. Cardiol. 51 (2008) 724–730. [13] H. Wada, T. Dohi, K. Miyauchi, et al., Prognostic impact of nutritional status assessed by the Controlling Nutritional Status score in patients with stable coronary artery disease undergoing percutaneous coronary intervention, Clin. Res. Cardiol. 106 (11) (2017) 875–883. [14] G. Basta, K. Chatzianagnostou, U. Paradossi, et al., The prognostic impact of objective nutritional indices in elderly patients with ST-elevation myocardial infarction undergoing primary coronary intervention, Int. J. Cardiol. 221 (2016) 987–992. [15] Q.J. Chen, H.J. Qu, D.Z. Li, et al., Prognostic nutritional index predicts clinical outcome in patients with acute ST-segment elevation myocardial infarction undergoing primary percutaneous coronary intervention, Sci. Rep. 7 (2017) 3285. [16] I. Nishi, Y. Seo, Y. Hamada-Harimura, et al., Nutritional screening based on the controlling nutritional status (CONUT) score at the time of admission is useful for longterm prognostic prediction in patients with heart failure requiring hospitalization, Heart Vessel. 32 (2017) 1337–1349. [17] Z. Zhang, S.L. Pereira, M. Luo, E.M. Matheson, Evaluation of blood biomarkers associated with risk of malnutrition in older adults: a systematic review and meta-analysis, Nutrients 9 (2017). [18] O. Bouillanne, G. Morineau, C. Dupont, et al., Geriatric Nutritional Risk Index: a new index for evaluating at-risk elderly medical patients, Am. J. Clin. Nutr. 82 (2005) 777–783. [19] K. Kalantar-Zadeh, J.D. Kopple, M.H. Humphreys, G. Block, Comparing outcome predictability of markers of malnutrition-inflammation complex syndrome in haemodialysis patients, Nephrol. Dial. Transplant. 19 (2004) 1507–1519. [20] J. Ignacio de Ulibarri, A. Gonzalez-Madrono, N.G. de Villar, et al., CONUT: a tool for controlling nutritional status. First validation in a hospital population, Nutr. Hosp. 20 (2005) 38–45. [21] M.R. Alvares-da-Silva, T. Reverbel da Silveira, Comparison between handgrip strength, subjective global assessment, and prognostic nutritional index in assessing malnutrition and predicting clinical outcome in cirrhotic outpatients, Nutrition 21 (2005) 113–117. [22] K. Fearon, J. Arends, V. Baracos, Understanding the mechanisms and treatment options in cancer cachexia, Nat. Rev. Clin. Oncol. 10 (2013) 90–99. [23] R. Aquilani, C. Opasich, M. Verri, et al., Is nutritional intake adequate in chronic heart failure patients? J. Am. Coll. Cardiol. 42 (2003) 1218–1223. [24] J.G. Carr, L.W. Stevenson, J.A. Walden, D. Heber, Prevalence and hemodynamic correlates of malnutrition in severe congestive heart failure secondary to ischemic or idiopathic dilated cardiomyopathy, Am. J. Cardiol. 63 (1989) 709–713. [25] L.M. Freeman, R. Roubenoff, The nutrition implications of cardiac cachexia, Nutr. Rev. 52 (1994) 340–347. [26] L.P. Fried, C.M. Tangen, J. Walston, et al., Frailty in older adults: evidence for a phenotype, J. Gerontol. A Biol. Sci. Med. Sci. 56 (2001) M146–56. [27] F. Cacciatore, P. Abete, F. Mazzella, et al., Frailty predicts long-term mortality in elderly subjects with chronic heart failure, Eur. J. Clin. Investig. 35 (2005) 723–730. [28] J.L. Purser, M.N. Kuchibhatla, G.G. Fillenbaum, T. Harding, E.D. Peterson, K.P. Alexander, Identifying frailty in hospitalized older adults with significant coronary artery disease, J. Am. Geriatr. Soc. 54 (2006) 1674–1681.