Red Cell Distribution Width and Mortality in Patients Undergoing Percutaneous Coronary Intervention

Red cell distribution width (RDW) can effectively predict prognosis in coronary artery disease (CAD) patients following percutaneous coronary intervention (PCI). There is currently no relevant research to demonstrate a linear or non-linear association between RDW and mortality. This is a multi-center, retrospective cohort study, with data collected from 2006 to 2017. Source data included electronic medical records of the Integrated Medical Database of National Taiwan University Hospital, and health insurance claims from the National Health Insurance Administration. Patients were stratified into five groups according to RDW values (13.4%, 14.1%, 14.8%, and 15.9%). Multivariable logistic and Cox regression analyses were used to determine 1-year all-cause and cardiovascular (CV) mortalities. Data of 10,669 patients were analyzed and those with the lowest RDW (≤13.3%) served as the reference group. The adjusted odds ratios (ORs) of 1-year all-cause mortality from the second to fifth RDW group were 1.386, 1.589, 2.090, and 3.192, respectively (p for trend < 0.001). The adjusted ORs of 1-year CV mortality were 1.555, 1.585, 1.623, and 2.850, respectively (p for trend = 0.015). The adjusted hazard ratios (HRs) of 1-year all-cause mortality were 1.394, 1.592, 2.003, and 2.689, respectively (p for trend = 0.006). The adjusted HRs of 1-year CV mortality were 1.533, 1.568, 1.609, and 2.710, respectively (p for trend = 0.015). RDW was an independent predicting factor and had a linear relationship with the 1-year all-cause and CV mortalities in patients undergoing PCI. Thus, RDW may be a clinically useful parameter to predict the mortality in those patients.


Introduction
Red cell distribution width (RDW) is the distribution of erythrocyte sizes derived from automated hematology analyzers, which can be used as a reliable index for anisocytosis [1]. RDW is calculated as a percentage value obtained by dividing a standard deviation of the erythrocyte size distribution by the mean red cell volume [2]. The normal value is between 11-15% [3], but it can be affected by inter-instrument differences including impedance or optical techniques and inter-laboratory differences [4]. In addition, RDW is affected by blood transfusion [5], acute or chronic heart failure [6], autoimmune disease [7] and neoplasms [8]. Currently RDW is mainly used as a differential diagnosis of microcytic or normocytic anemia. However, recent research has found that RDW can be used as an index for risk stratification of cardiovascular (CV) diseases [9], and it can also improve the effectiveness of mortality stratification for patients with high-risk critical illness [10]. RDW may also be applicable to various other populations. Whether RDW is a specific marker of risk-stratification still needs to be evaluated.
Coronary artery disease (CAD) is caused by coronary artery atherosclerosis and endothelial hyperplasia [11]. Plaque rupture of the coronary artery can cause thrombosis, and block the coronary blood flow, which leads to myocardial infarction (MI) and necrosis [11]. The diagnostic standard for CAD is coronary angiography [12,13]. In addition to the control of risk factors such as hypertension, diabetes, hyperlipidemia, and smoking, invasive percutaneous coronary intervention (PCI) for CAD is the standard treatment [12,13]. With the evolution of technology, such as the development of balloon angioplasty, bare-metal stents and drug-eluting stents, PCI can effectively reduce the mortality rate in patients with ST elevation MI [14]. Many studies including meta-analysis showed that RDW can effectively predict all-cause mortality and CV mortality in patients following PCI [15][16][17][18]. However, because of the insufficient sample size, previous studies only use one specific value of RDW to analyze the differences between the lower and higher groups, and the cut-off values of RDW identified in studies in the meta-analysis are also different. There is no relevant research to demonstrate whether the relationship between RDW and mortality is a linear or non-linear association. The purpose of this study is to conduct a large-scale multi-center retrospective study, spanning more than 10 years, to analyze the role of RDW in the prediction of 1-year all-cause mortality and CV mortality in CAD patients undergoing PCI.

Study Design and Patients Selection
This study is a multi-center and retrospective cohort study. Eligible patients were considered to meet the following criteria: (1) age ≥ 20 years; (2) patients who received PCI from 2006 through 2017. Exclusion criteria were: (1) no RDW data when admission; (2) patients who died during the index hospitalization. For patients hospitalized more than once during the study period, the first hospitalization was used as the index hospitalization. Then, patients were stratified into 5 groups, according to RDW values on admission [10].

Data Sources
This study used the electronic medical records incorporated into the Integrated Medical Database of National Taiwan University Hospital (NTUH-iMD), including the data from National Taiwan University Hospital (NTUH) from 2007 onward, the NTUH Yunlin Branch from February 2014 onward, and the NTUH Hsinchu Branch from November 2014 onward. In addition, health insurance claims data from Taiwan National Health Insurance (NHI) Administration and the national mortality data were utilized. Taiwan has implemented universal health insurance since 1995 [19], and approximately 97% of citizens have participated in the insurance [20]. Relevant clinical data including personal identification number, date of birth, gender, date of outpatient visit, hospital admission and discharge, diagnostic codes, procedure codes, and drugs administered, are available from the NHI for research. Through the end of 2015, the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) system was used as the diagnostic codes and the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) system has been used since 2016.
In this study, patients with PCI in the inpatient procedure were identified from the NTUH-iMD. Then, the patients' identification number was used to link the database of NHI Administration to obtain background characteristics of patients and national mortality data to obtain the definite date of death. All linkages were carried out under guidelines specified by the Department of Statistics, Ministry of Health and Welfare and National Taiwan University Hospital to protect individuals' privacy. This research was approved by Institutional Review Board of NTUH (201802076RINB), and all research methods and data analysis were carried out in accordance with regulations.

Clinical Attributes and Outcomes Assessment
Baseline characteristics including age, sex, body mass index, type and total number of coronary stents, diagnosis of hospitalization, medications at discharge, and laboratory data of index hospitalization were ascertained from the NTUH-iMD. Previous medical history was assessed from both the NTUH-iMD and the Taiwan NHI claims database. The evaluation time period of the baseline characteristics was 18 months prior to the index hospitalization. In addition to comorbidity defined in the Elixhauser comorbidity index [21], we also defined hyperlipidemia as ICD9: 272. ICD10: I20.1x, I20.8x, I20.9x, I25.x. A patient with diagnosis codes of interest two or more times in the outpatient records during the 18-month baseline period, or once in the inpatient records, was defined as having a comorbidity of interest. The clinical outcomes of this study included the all-cause mortality and CV mortality (ICD9: 353.x-459.x; ICD10: I05.x-I99.x) at 6 months and 1 year after discharge from the index hospitalization, respectively.

Statistical Analyses
The first RDW value after admission was used to categorize the patients into five groups according to previously reported cut-off values (13.4%, 14.1%, 14.8%, and 15.9%) [10]. Mean and standard deviation were used to describe the continuous variables. Number and percent were used to present the categorical data. ANOVA and the chi-squared test were used to compare the characteristics between RDW groups. A multivariable logistic regression model, adjusting for the confounding variables, was used to compare the various clinical outcomes between RDW groups. A receiver operating characteristic (ROC) curve was plotted to determine the best cut-off value of RDW, according to the maximum value of Youden index [22]. A Cox proportional hazard regression analysis was also used for time-to-event analysis and the cumulative probabilities of all-cause or CV mortalities were estimated by the Kaplan-Meier curves. All analysis was performed using SAS software, version 9.4 (SAS Institute, Inc., Cary, NC, USA).

Study Subjects
We identified 15,319 patients undergoing PCI from January 2006 to December 2017, 4405 patients had no RDW data during hospitalization and 245 patients died in the hospital, resulting in 10,669 survivors discharged from the hospitals after PCI; 8382 (78.6%) were men and 2287 (21.4%) were women, and the mean age was 65.4 ± 12.1 years (Supplementary Figure S1). We divided the patients into five groups, according to the first value of the RDW during the hospitalization. There were significant differences in basic characteristics between the RDW groups, including: age; sex; hemoglobin; estimated glomerular filtration rate; white blood cells; total cholesterol; triglyceride; body mass index; hypertension; hyperlipidemia; diabetes mellitus; current smoking status; previous MI previous ischemic stroke; peripheral vascular disease; chronic pulmonary disease; previous PCI; previous coronary artery bypass grafting surgery; admission diagnosis; and type of intervention. Moreover, they also used different medications, including: ticagrelor; amiodarone; diltiazem; dihydropyridine calcium channel blockers; angiotensin-converting enzyme inhibitors; loop diuretics; spironolactone; statins; oral antidiabetic drugs; insulin; and proton-pump inhibitors. The patients in higher RDW groups had lower hemoglobin and estimated glomerular filtration rate, and a higher rate of hypertension, diabetes, previous MI, previous ischemic stroke, previous coronary artery bypass grafting surgery, and myocardial infarction. Therefore, the patients in the higher RDW groups were sicker than patients in the lower RDW groups (Table 1).

Multivariable Logistic Regression Analysis
The lowest RDW (≤13.3%) group was used as the reference group. The crude odd ratios (   Table 3). The ORs of 6-month mortalities showed similar results, and are shown in Table 3.

Cox Proportional Hazards Regression Analysis and Kaplan-Meier Curves
The lowest RDW (≤13.3%) group was used as the reference group. The crude hazard ratios (HRs) Table S1). The cumulative all-cause death and CV death were estimated by Kaplan-Meier plots. The log-rank test showed significant differences between the groups (Figure 1). Biomedicines 2021, 9, x FOR PEER REVIEW 7 of 12 respectively (p for trend = 0.015) (Supplementary Table S1). The cumulative all-cause death and CV death were estimated by Kaplan-Meier plots. The log-rank test showed significant differences between the groups (Figure 1).

Discussion
In this large-scale real-world study from a non-Caucasian population to evaluate the relationship between RDW and mortality in CAD patients undergoing PCI, we found that RDW was an independent factor associated with 1-year all-cause mortality and CV mortality in CAD patients undergoing PCI. RDW had a positive linear and dose-response relationship with 1-year all-cause mortality and CV mortality. The results indicated that the higher the RDW, the higher the all-cause mortality and CV mortality in patients with CAD undergoing PCI.
Previous studies have shown that RDW was associated with CV diseases [23] and RDW could be used as a parameter to predict all-cause mortality and CV mortality [24]. The possible mechanism is that RDW is an inflammation marker [25]. Inflammation can lead to abnormal bone marrow function, resulting in poor efficiency of red blood cell production, and also affects RBC membrane permeability, which causes reticulocytes to enter the peripheral blood circulation, and increases the proportion of immature RBC, leading to an increase in RDW [26]. Inflammatory markers are associated with mortality and major adverse CV events [27,28]. Therefore, RDW may be used as an indicator of mortality prediction. Another possible mechanism is that anisocytosis is directly involved in the

Discussion
In this large-scale real-world study from a non-Caucasian population to evaluate the relationship between RDW and mortality in CAD patients undergoing PCI, we found that RDW was an independent factor associated with 1-year all-cause mortality and CV mortality in CAD patients undergoing PCI. RDW had a positive linear and dose-response relationship with 1-year all-cause mortality and CV mortality. The results indicated that the higher the RDW, the higher the all-cause mortality and CV mortality in patients with CAD undergoing PCI.
Previous studies have shown that RDW was associated with CV diseases [23] and RDW could be used as a parameter to predict all-cause mortality and CV mortality [24]. The possible mechanism is that RDW is an inflammation marker [25]. Inflammation can lead to abnormal bone marrow function, resulting in poor efficiency of red blood cell production, and also affects RBC membrane permeability, which causes reticulocytes to enter the peripheral blood circulation, and increases the proportion of immature RBC, leading to an increase in RDW [26]. Inflammatory markers are associated with mortality and major adverse CV events [27,28]. Therefore, RDW may be used as an indicator of mortality prediction. Another possible mechanism is that anisocytosis is directly involved in the pathogenic process of CV diseases [29]. The injury of the fibrous cap of the atherosclerosis plaque causes thrombosis, and red blood cells are entrapped in the atherosclerosis plaque of blood vessel wall. Once red blood cells are entrapped in the atherosclerosis plaque, a series of inflammatory reaction processes are activated and lead to CV diseases [29].
Pilling et al. reported a large-scale observational study including 240,477 healthy volunteers from UK Biobank [23]. The study demonstrated that those with higher RDW had higher all-cause mortality, with linear dose-response relationship, and also predicted new onsets of wide common conditions including incident coronary artery disease, heart failure, peripheral artery disease, atrial fibrillation, stroke and cancer. Aside from healthy volunteers, RDW has also been reported to be a robust predictor of the risk of all-cause mortality in older adults [30], the critically ill [10,31], non-CV critically patients [32], and those with heart failure [18,33], chronic kidney disease [34], and cancers including gliomas [35], and gastrointestinal cancers [36,37]. This evidence indicates that RDW may be a marker for predicting mortality from various diseases.
A few studies have shown that RDW can be used as a predictor of mortality in CAD patients after PCI [38][39][40][41][42][43]. However, previous studies only dichotomize RDW groups. The reported cut-off values ranged from 13.1% to 14.8%, and the studies showed a higher risk of all-cause mortality and CV mortality in high RDW group, in comparison with low RDW groups [38,39,44]. Bao et al. [16] reported a meta-analysis of 12 studies with a total of 17,113 CAD patients undergoing PCI. The study demonstrated the risk ratio was 1.77 for all-cause mortality and 1.70 for CV mortality, comparing the higher RDW group with the lower RDW group [16]. Latif et al. [15] reported another meta-analysis including 21 studies and 56,425 CAD patients undergoing PCI. The results showed that higher RDW had higher in-hospital all-cause mortality (OR 2.41), long-term all-cause mortality (OR 2.44) and cardiac mortality (OR 2.65) [15]. The aforementioned two meta-analyses conducted a quantitative synthesis of smaller studies with high heterogeneity. Different cut-off values of RDW may be potential sources of this heterogeneity. Our research included more than 10,000 patients from three hospitals, with divided RDW into five groups; the result showed a linear and dose-response relationship between RDW and mortality risks. Thus, our study demonstrated clearly that RDW was a robust dose-response predictor of all-cause mortality and CV mortality in CAD patients undergoing PCI.
Our study had some limitations. Firstly, this study consisted of hospital-based research, including one medical center and two regional hospitals in Taiwan. Whether the findings of this study could be extrapolated to other hospital levels, or patients in other countries, still needs to be evaluated. However, NHI and mortality follow-up with approximately 97% [20] participation of the citizens, mitigating the limitations of the hospital-based data source. Secondly, this study did not distinguish between acute and chronic context of coronary artery disease, and could not show whether the two populations had different results. However, this study used multivariable analysis to adjust the diagnosis when admission included acute or chronic coronary artery disease. RDW is still an independent and effective factor related to mortality. Thirdly, this study spanned more than 10 years, and the standard practice of medical care might change overtime. Fourthly, nutritional information in the study population might affect the value of RDW [5]. This study lacked the adjustment of some unmeasured confounding factors, such as iron, folate, vitamin B12 and other nutritional information. Fifthly, this study did not collect information of inflammatory biomarkers such as C-reactive protein [27], procalcitonin [45] or interleukin-6 [28], as these biomarkers were not checked in most of our study subjects. Finally, this study only discussed baseline RDW on admission, and did not include subsequent changes in RDW for analysis.

Conclusions
We found a linear and dose-response relationship between RDW and all-cause and CV mortalities in patients with CAD undergoing PCI. Thus, RDW may be used as a clinical parameter to predict future mortality and prognosis in patients with CAD undergoing PCI.