Journal of Clinical Medicine Article Major Adverse Cardiovascular Events in Coronary Type 2 Diabetic Patients: Identification of Associated Factors Using Electronic Health Records and Natural Language Processing Carlos González-Juanatey 1, * , Manuel Anguita-Sánchez 2 , Vivencio Barrios 3 , Iván Núñez-Gil 4 , Juan José Gómez-Doblas 5 , Xavier García-Moll 6 , Carlos Lafuente-Gormaz 7 , María Jesús Rollán-Gómez 8 , Vicente Peral-Disdier 9 , Luis Martínez-Dolz 10 , Miguel Rodríguez-Santamarta 11 , Xavier Viñolas-Prat 6 , Toni Soriano-Colomé 12 , Roberto Muñoz-Aguilera 13 , Ignacio Plaza 14 , Alejandro Curcio-Ruigómez 15 , Ernesto Orts-Soler 16 , Javier Segovia 17 , Víctor Fanjul 18 , Ángel Cequier 19 and SAVANA Research Group † 1 Hospital Universitario Lucus Augusti, 27003 Lugo, Spain 2 Hospital Universitario Reina Sofía, Instituto Maimonides de Investigación Biomédica de Córdoba (IMIBIC), Universidad de Córdoba, 14014 Córdoba, Spain 3 Hospital Universitario Ramón y Cajal, 28034 Madrid, Spain 4 Hospital Clínico Universitario San Carlos, 28040 Madrid, Spain 5 Hospital Universitario Virgen de la Victoria CIBERCV (Centro de Investigación Biomédica en Red Enfermedades Cardiovasculares) and IBIMA (Instituto de Investigación Biomédica de Málaga), 29010 Málaga, Spain 6 Hospital Universitario Santa Creu i Sant Pau, 08041 Barcelona, Spain Citation: González-Juanatey, C.; 7 Hospital Universitario de Albacete, 02006 Albacete, Spain Anguita-Sánchez, M.; Barrios, V.; 8 Hospital Universitario Río Hortega, 47012 Valladolid, Spain Núñez-Gil, I.; Gómez-Doblas, J.J.; 9 Hospital Universitario Son Espases, 07120 Palma de Mallorca, Spain García-Moll, X.; Lafuente-Gormaz, C.; 10 Hospital Universitario y Politécnico La Fe, CIBERCV, IIS La Fe, 46026 València, Spain 11 Hospital Universitario de León, 24071 León, Spain Rollán-Gómez, M.J.; Peral-Disdier, V.; 12 Hospital Vall d’Hebron, CIBERCV, 08035 Barcelona, Spain Martínez-Dolz, L.; et al. Major 13 Hospital Infanta Leonor, 28031 Madrid, Spain Adverse Cardiovascular Events in 14 Hospital Infanta Sofía, 28703 Madrid, Spain Coronary Type 2 Diabetic Patients: 15 Hospital Universitario de Fuenlabrada, 28942 Madrid, Spain Identification of Associated Factors 16 Hospital General Universitario de Castellón, 12004 Castellón de la Plana, Spain Using Electronic Health Records and 17 Hospital Universitario Puerta de Hierro, 28222 Madrid, Spain Natural Language Processing. J. Clin. 18 Savana Research, 28013 Madrid, Spain Med. 2022, 11, 6004. https://doi.org/ 19 Hospital Universitario de Bellvitge, IDIBELL (Instituto de Investigación Biomédica de Bellvitge), Universidad 10.3390/jcm11206004 de Barcelona, 08007 Barcelona, Spain * Correspondence: carlos.gonzalez.juanatey@sergas.es; Tel.: +34-982-296-809 Academic Editors: Alessandro † SAVANA Research Group are shown in the Acknowledgements. Delitala and Francesco Giallauria Received: 8 September 2022 Abstract: Patients with Type 2 diabetes mellitus (T2DM) and coronary artery disease (CAD) are Accepted: 6 October 2022 at high risk of developing major adverse cardiovascular events (MACE). This is a multicenter, Published: 11 October 2022 retrospective, and observational study performed in Spain aimed to characterize these patients Publisher’s Note: MDPI stays neutral in a real-world setting. Unstructured data from the Electronic Health Records were extracted by with regard to jurisdictional claims in EHRead® , a technology based on Natural Language Processing and machine learning. The association published maps and institutional affil- between new MACE and the variables of interest were investigated by univariable and multivariable iations. analyses. From a source population of 2,184,662 patients, we identified 4072 adults diagnosed with T2DM and CAD (62.2% male, mean age 70 ± 11). The main comorbidities observed included arterial hypertension, hyperlipidemia, and obesity, with metformin and statins being the treatments most frequently prescribed. MACE development was associated with multivessel (Hazard Ratio Copyright: © 2022 by the authors. (HR) = 2.49) and single coronary vessel disease (HR = 1.71), transient ischemic attack (HR = 2.01), Licensee MDPI, Basel, Switzerland. heart failure (HR = 1.32), insulin treatment (HR = 1.40), and percutaneous coronary intervention This article is an open access article (PCI) (HR = 2.27), whilst statins (HR = 0.73) were associated with a lower risk of MACE occurrence. distributed under the terms and In conclusion, we found six risk factors associated with the development of MACE which were conditions of the Creative Commons Attribution (CC BY) license (https:// related with cardiovascular diseases and T2DM severity, and treatment with statins was identified as creativecommons.org/licenses/by/ a protective factor for new MACE in this study. 4.0/). J. Clin. Med. 2022, 11, 6004. https://doi.org/10.3390/jcm11206004 https://www.mdpi.com/journal/jcm J. Clin. Med. 2022, 11, 6004 2 of 12 Keywords: diabetes mellitus; coronary artery disease; MACE; risk factors; electronic health records; natural language processing 1. Introduction Diabetes mellitus (DM) is a key public health problem worldwide associated with the development of complications such as cardiovascular comorbidities, which include coronary artery disease (CAD) [1–3]. It was listed as the underlying cause of mortality in >85,000 deaths in the United States in 2019 [2], consisting mainly of cardiovascular complications and CAD development, the main causes of death and disability in patients with type 2 diabetes mellitus (T2DM) [4]. In addition, DM has been associated with a 2 to 4-fold increased risk of death due to heart disease [5]. Given the clinical burden that cardiovascular complications have on patients with T2DM, there has been an increased focus on the joint management of both entities, T2DM and cardiovascular disease (CVD). Furthermore, during the last decade, there has been increasing pressure from regulatory agencies for antidiabetic treatments to demonstrate cardiovascular safety and benefits, especially for major adverse cardiovascular events (MACE) [6]. However, patients included in controlled trials are generally selected on the basis of protocol eligibility and, therefore, do not fully represent the entire population encountered in clinical practice [7]. In this regard, real-world-based studies are needed to improve diagnostic and therapeutic approaches to palliate the burden of disease by reducing cardiovascular events. The present study aimed to use population-based data to provide real-world insights into the clinical characterization and treatment management of diabetic patients who develop CAD in Spain. Moreover, this study investigated the probability of developing MACE after presentation of CAD in T2DM patients and the potential risk or protective factors associated with its occurrence within this population. To do so, we used EHRead® (Madrid, Spain), a technology that applies Natural Language Processing (NLP) and machine learning, to extract, organize, and analyze the unstructured clinical information that health professionals register in patients’ electronic health records (EHRs) [8–13]. 2. Materials and Methods 2.1. Study Design This was a real-world, multicenter, retrospective, and observational study (ACORDE: “Assessment of medical management in CORonary DiabEtic Type 2 patients at high risk of cardiovascular events”) based on the secondary use of the unstructured data captured in the EHRs. A cross-sectional analysis of all patients was performed at the index date, defined as the first time in the study period when the patient fulfilled all inclusion and no exclusion criteria. At this point, demographics, comorbidities, vital signs, general characteristics of T2DM and CAD and treatments were evaluated. The cumulative incidence of MACE, defined as previously described by the presence of myocardial infarction, stroke, hospi- talization for unstable angina, and urgent coronary revascularization [14], was analyzed during the follow-up. Although included in this definition, all-cause or cardiovascular death was not analyzed as, due to the nature of the data source, their occurrence was not accurately identifiable. The follow-up period was defined as the period of time spanned from index date to the last EHR available for each patient within the study period. 2.2. Data Source The data source was the free-text information within the EHRs of 12 representative hospitals from six major regions in Spain: Madrid (Hospital Universitario de Fuenlabrada, Hospital Universitario Infanta Sofía, Hospital Universitario Infanta Leonor, Hospital Uni- versitario Puerta de Hierro), Catalonia (Hospial Universitari Vall d’Hebron, Hospital de la Santa Creu i Sant Pau), Valencia (Hospital Universitari i Politècnic La Fe, Hospital General J. Clin. Med. 2022, 11, 6004 3 of 12 Universitario de Castellón), Balearic Islands (Hospital Universitari Son Espases), Castilla La Mancha (Complejo Hospitalario Universitario de Albacete), and Castilla y León (Hospi- tal Universitario Río Hortega, Hospital Universitario de León). Data entry in outpatient clinical reports, discharge reports, emergency reports, prescriptions, and other medical reports, were collected from all available services and departments in each participating site, including inpatient hospital, outpatient hospital, and emergency room, between 1 January 2014 and 31 December 2018. 2.3. Study Subjects The source population of the study comprised all adult patients with available EHRs in the participating hospitals within the study period with at least 6 months of follow-up. Patients with T2DM and a diagnosis of CAD were included. These two entities were considered when they were contained in the unstructured, free-text information in the EHRs based on clinical diagnosis. T2DM were also considered when documented ongoing use of glucose-lowering drugs (oral hypoglycemic agents) was found for at least 6 months. CAD was considered if there was evidence of stenosis ≥ 50% of at least 1 coronary artery but without a history of having a previous myocardial infarction or stroke, and without planned coronary, cerebrovascular, or peripheral arterial revascularization. Patients with prior myocardial infarction or stroke, history of liver cirrhosis or liver cancer, intracranial bleeding, renal failure requiring dialysis, or ongoing treatment with anticoagulant medica- tion at index date or non-available follow-up information spanning at least 6 months were excluded from the study. 2.4. Extraction of the Unstructured Free-Text from EHRs Unstructured clinical data were extracted using the EHRead® technology, which uses NLP, machine learning, and deep learning techniques for extracting free text from de- identified and processed EHR and translates it into a study database. EHRead® technology have been described elsewhere [9,10,12,13,15–19]. Using the information obtained from this process, a statistical model was generated to describe the demographic and clinical characteristics of the population with T2DM and CAD. Information was analyzed and expressed by means of concepts that contain the most significant information in the text. The terminology considered by EHRead® includes codes, concepts, synonyms, and definitions used in clinical documentation and is based on SNOMED CT [20,21]. 2.5. External Validation of EHRead® Performance EHRead® performance was assessed to evaluate the ability of the technology to identify mentions of coronary disease and its related clinical variables within patients’ EHRs [9,12,13,22]. First, a comparison between a physician-annotated set of EHRs (i.e., “gold standard”) and the “output” of EHRead® upon reading that same set of EHRs was car- ried out independently at each hospital site [9]. Thus, two designated expert physicians from each hospital annotated a set of randomly selected records while a third physician reviewed the annotations made and resolved any possible discrepancies among the two site physicians. The “output vs. gold standard” comparison was reflected in the standard evaluation metrics, including precision, recall, and their harmonic mean (F1-score). As previously described, we demonstrated a high performance of EHRead® at identifying records that contain mentions of “Coronary disease” in the target population with a pre- cision, recall, and F1-score of 0.89, 0.73, and 0.80, respectively. Moreover, the evaluation yielded a F1-score of above 90% in most of the variables analyzed [9,10]. 2.6. Statistical Data Analyses Categorical variables were described via frequency tables; whereas numerical variables were presented using summary tables that included the mean, standard deviation (SD), median, and quartiles (Q1, Q3). Missing data were handled according to the nature of the J. Clin. Med. 2022, 11, 6004 4 of 12 data collection process, thus lack of information (i.e., unavailable data in patients’ EHRs) was considered a “true zero” for binary variables (e.g., absence of a comorbidity) but was treated as missing data for numerical variables (e.g., laboratory values). No missing data imputation was performed. Cumulative incidence of new cases of MACE was estimated through the Kaplan–Meier approach. Both univariable and multivariable analyses were performed to investigate the association between the occurrence of new MACE and the variables of interest. Variables with >20% missing values or with zero variance were excluded. In the univariable analysis, a Cox proportional hazard (PH) regression model was fitted to the study population for each variable at baseline. In the multivariable approach, a data-driven variable and model selection was performed. First, a Cox PH model was fitted to the study population using all eligible study variables. Then, we carried out backward stepwise selection (i.e., the least informative variables were excluded one at a time, and new models were fitted in a recursive manner). We used the Bayesian information criterion (BIC) to reach a model with the optimal explanatory variables. Significant differences were considered when p < 0.05. All analyses were performed using “R” software, version 4.0.2 (The R Foundation for Statistical Computing, Vienna, Austria). 2.7. Ethical Considerations and Study Approval This study was classified as a “non-post-authorization study” by the Spanish Agency of Medicines and Health Products (AEMPS) and was approved by the Institutional Review Board (IRB) of each participating hospital. All methods and analyses were conducted in compliance with local legal and regulatory requirements, as well as generally accepted research practices described in the Helsinki Declaration in its latest edition, and Good Pharmacoepidemiology Practices. Data were analyzed from de-identified EHRs, which were aggregated in an irreversibly dissociated manner. Thus, individual patient consent was not required in the study. 3. Results 3.1. Study Population The source population of the study comprised a total of 2,184,662 adult patients with available EHRs in the hospitals of interest including 217,632 (10%) patients with a confirmed diagnosis of T2DM. The study population only included adult patients with T2DM and stable CAD diagnosis with no previous history of myocardial infarction and stroke, which yielded to a final population of 4072 patients. In total, 37.4% of the patients presented insulin-dependent diabetes. Type of CAD was identified in 60% of the patients. Of which, 833 of the patients (34.21% of the patients with CAD type identified within the EHR) had single coronary vessel disease and 1,602 (65.79%) multivessel coronary disease. Patients were followed up for a median of 33.6 months (IQ1-3: 19.5–47.1). 3.2. Demographic and Clinical Characteristics The sociodemographic and comorbidities of the study population at index date are summarized in Table 1 and Figure 1. As shown, the mean age of the patients was 70 ± 11 years, with two thirds being male (62.2%) and 17.5% current smokers (Table 1). Table 1. Sociodemographic of the study population at index. Study Population (N = 4072) Age, years Mean (SD) 70 (11) Median (Q1,Q3) 71 (63,78) J. Clin. Med. 2022, 11, x FOR PEER REVIEW 5 of 12 J. Clin. Med. 2022, 11, 6004 5 of 12 Table 1. Sociodemographic of the study population at index Study Population Table 1. Cont. (N = 4072) Age, years Study Population Mean (SD) 70 (N = 4072) (11) Gender, n (%)(Q1,Q3) Median 71 (63,78) Gender, n (%) Female 1533 (37.6) Female Male 15332531 (37.6) (62.2) Maleuse Tobacco 25312208 (62.2) (54.2) Tobacco use Current smoker 2208 713 (54.2) (17.5) Current smoker 713 (17.5) Former smoker 1495 (36.7) Former smoker 1495 (36.7) SD: Standard deviation. SD: Standard deviation. Figure 1.1. Comorbidities Figure Comorbidities of of the the study study population population at at index. index. (A) Cardiovascular and (A) Cardiovascular and (B) (B) non- non- cardiovascularcomorbidities cardiovascular comorbiditiesare areshown. shown.CKD: CKD:Chronic Chronickidney kidneydisease; disease;COPD: COPD:Chronic Chronicobstructive obstructive pulmonary disease. pulmonary disease. The most The most common common cardiovascular cardiovascular comorbidities comorbidities reported reported in in the the target targetpopulation population includedarterial included arterialhypertension hypertension (84.65%), (84.65%), and and heart heartvalve valvedisease disease(38.51%) (38.51%) (Figure (Figure1A). 1A).In In regardto regard toother othercomorbidities comorbiditiesobserved, observed, the themost mostfrequent frequentdisorder disorderwaswashyperlipidemia hyperlipidemia (40.91%),followed (40.91%), followedbybyobesity obesity(32.60%) (32.60%)(Figure (Figure1B). 1B). As shown in Table S1, our study population As shown in Table S1, our study population had hadclinical clinicalparameters parametersconsistent consistentwith with metabolic abnormalities including elevated glucose, body mass index, lipids metabolic abnormalities including elevated glucose, body mass index, lipids parameters parameters andhemoglobin and hemoglobinA1cA1cvalues. values. J. Clin. Med. 2022, 11, x FOR PEER REVIEW 6 of 12 J. Clin. Med. 2022, 11, 6004 6 of 12 3.3. 3.3. Disease Disease Management Management Treatment strategies in Treatment strategies in our our study study population population at at index index date date included included pharmacological pharmacological management management and and interventional interventional approaches. approaches. Data regardingpharmacological Data regarding pharmacological treatment treatment for thefor the management management of patientsofdiagnosed patients diagnosed with T2DM and CAD are with T2DM and CAD are shown in Figure 2. shown in Figure 2. Figure Figure 2. 2. Pharmacological Pharmacological treatment treatment of of the the study study population population at at index. index. ACE: ACE:Angiotensin-converting Angiotensin-converting enzyme; ARB: Angiotensin II receptor blockers; GLP1-RA: Glucagon-like peptide-1 receptor enzyme; ARB: Angiotensin II receptor blockers; GLP1-RA: Glucagon-like peptide-1 receptor agonists; agonists; iDPP4: Dipeptidil peptidasa-4 inhibitors; iSGLT2: Sodium-glucose cotransporter-2; iDPP4: Dipeptidil peptidasa-4 inhibitors; iSGLT2: Sodium-glucose cotransporter-2; NOAC: Non- NOAC: Non-vitamin K antagonist oral anticoagulants; VKA: Vitamin K antagonists. vitamin K antagonist oral anticoagulants; VKA: Vitamin K antagonists. Metformin was identified as the most frequently prescribed antidiabetic drug Metformin was identified as the most frequently prescribed antidiabetic drug (77.6%), (77.6%), followed by insulin/analogs (25%), sulfonylureas (21.64%), dipeptidyl peptidase- followed by insulin/analogs (25%), sulfonylureas (21.64%), dipeptidyl peptidase-4 in- 4 inhibitors (iDPP4) (20.83%), and glinides (12.45%). Other treatments such as statins hibitors (iDPP4) (20.83%), and glinides (12.45%). Other treatments such as statins (79.15%) (79.15%) and antiplatelet and antiplatelet therapy, therapy, mainly ASAmainly (64.0%),ASA were (64.0%), commonly were commonly prescribed prescribed according to the according to the available information available information within the EHRs. within the EHRs. Regarding interventionalprocedures, Regarding interventional procedures, wewe observed observed thatthat a total a total ofpatients of 1574 1,574 patients (38.8%) (38.8%) were revascularized with percutaneous coronary interventions (PCI) and 585 and were revascularized with percutaneous coronary interventions (PCI) 585 patients patients (14.4%) (14.4%) were werewith treated treated with coronary coronary arterygraft artery bypass bypass graft (CABG). (CABG). 3.4. Major Adverse Cardiovascular Events (MACE) J. Clin. Med. 2022, 11, 6004 7 of 12 Med. 2022, 11, x FOR PEER REVIEW 7 of 12 3.4. Major Adverse Cardiovascular Events (MACE) We evaluated the Wecumulative evaluated theincidence of MACE, cumulative during incidence the study of MACE, follow-up. during the study follow-up. As shown in Table 2, the in As shown probability of developing Table 2, the at least probability of one MACE developing after at least one4-years MACE after 4-years follow-up wasfollow-up 27.09%, with it being more probable to undergo myocardial infarction was 27.09%, with it being more probable to undergo myocardial infarction (13.89%) than (13.89%) developingthanischemic strokeischemic developing (6.94%),stroke unstable angina (6.94%), (4.54%),angina unstable or urgent (4.54%), or urgent coronary revascularization (8.69%). coronary revascularization (8.69%). Table 2. Cumulative incidence of MACE. Table 2. Cumulative incidence of MACE. 1-Year FU 2-Year FU 3-Year FU 4-Year FU 1-Year FU 2-Year FU 3-Year FU 4-Year FU MACE * 10.28 16.32 21.18 27.09 MACE * 10.28 16.32 21.18 27.09 Myocardial infarction 4.05 7.8 10.79 13.89 Myocardial infarction 4.05 7.8 10.79 13.89 Ischemic stroke 1.77 3.41 5.05 6.94 Unstable anginaIschemic stroke 1.67 1.77 2.49 3.41 3.45 5.05 4.54 6.94 Unstable angina Urgent revascularization 4.04 1.67 5.57 2.49 6.82 3.45 8.69 4.54 FU: Follow-up; MACE: Major Urgent adverse cardiovascular revascularization 4.04 events. 5.57 6.82 8.69 * Cumulative incidence of new cases of MACE was estimated through the FU: Follow-up; MACE: Major adverse cardiovascular events. Kaplan-Meier approach. Each value represents the probability of MACE at different FU time points. * Cumulative incidence of new cases of MACE was estimated through the Kaplan-Meier approach. Each value represents the probability of MACE at different FU time points. Cumulative incidence curves of the first MACE occurrence after study entry and detailed probabilityCumulative of individualincidence componentscurves of the over of interest first study MACEfollow-up occurrenceareafter shown study entry and detailed probability in Figures 3 and S1, respectively. of individual components of interest over study follow-up are shown in Figure 3 and Figure S1, respectively. Figure 3. Cumulative Figureincidence of MACE 3. Cumulative over time incidence duringover of MACE the follow-up period. time during Probability the follow-up of any period. Probability of any new MACE (myocardial infarction, stroke, unstable angina, and urgent revascularization) is shown. new MACE (myocardial infarction, stroke, unstable angina, and urgent revascularization) is shown. The number of patients at risk across the follow-up period is indicated below. The number of patients at risk across the follow-up period is indicated below. The univariableThe analysis using Cox univariable PH models analysis showed using Cox that age, PH models historythat showed of smoking, age, history of smoking, and the time ofand firstthe mention of T2DM and CAD disease in the EHRs were significantly time of first mention of T2DM and CAD disease in the EHRs were significantly associated withassociated the occurrence of new with the MACE of occurrence during the follow-up new MACE during period, while female the follow-up period, while female gender was inversely associated with the development of MACE (p < 0.001) (Table S2). gender was inversely associated with the development of MACE (p < 0.001) (Table S2). Regarding comorbidities, in general, the presence of all cardiovascular disorders was associated with the occurrence of the first MACE detected during the follow-up. Likewise, J. Clin. Med. 2022, 11, 6004 8 of 12 J. Clin. Med. 2022, 11, x FOR PEER REVIEW 8 of 12 Regarding comorbidities, in general, the presence of all cardiovascular disorders was associated with the occurrence of the first MACE detected during the follow-up. Likewise, diabetic retinopathy, hypothyroidism, and COPD/asthma were found to be associated diabetic retinopathy, hypothyroidism, and COPD/asthma were found to be associated with the appearance of new MACE in the study (Table S2). In addition, new MACE were with the appearance of new MACE in the study (Table S2). In addition, new MACE were associated with prescription of insulin (p < 0.001), some hypoglycemic agents including associated with prescription of insulin (p < 0.001), some hypoglycemic agents including sulfonylureas (p = 0.008) and alpha-glucosidase (p = 0.049), anticoagulant and antiplatelet sulfonylureas (p = 0.008) and alpha-glucosidase (p = 0.049), anticoagulant and antiplatelet therapy (p = 0.021 and p < 0.001, respectively), beta blockers, calcium channel, nitrates, therapy (p = 0.021 and p < 0.001, respectively), beta blockers, calcium channel, nitrates, ranolazines, lipid lowering drugs (p < 0.001 each), diuretics (p = 0.012), and statins (p = ranolazines, lipid lowering drugs (p < 0.001 each), diuretics (p = 0.012), and statins (p = 0.003) 0.003) (Table S2). Revascularization treatment, including PCI and CABG were also found (Table S2). Revascularization treatment, including PCI and CABG were also found to be to be related the occurrence of new MACE during the study period (both p < 0.001). related the occurrence of new MACE during the study period (both p < 0.001). InIna amultivariable multivariableapproach approachusing usingbackward backwardstepwise stepwiseselection selectionbased basedon onBIC, BIC,we we found 7 independent variables to be associated with MACE during found 7 independent variables to be associated with MACE during follow-up, including follow-up, including coronaryinvolvement coronary involvement with with multivessel multivessel andandsingle singlecoronary coronaryvessel vesseldisease disease(HR (HR= 2.49; 95% = 2.49; CI = 2.01–3.10; p < 0.001), and (HR 1.71; 95% CI = 1.34–2.18; p < 0.001), 95% CI = 2.01–3.10; p < 0.001), and (HR 1.71; 95% CI = 1.34–2.18; p < 0.001), transient transient ischemic attack (HR ischemic = 2.01; attack (HR95% = 2.01;CI =95% 1.46–2.77; p < 0.001), CI = 1.46–2.77; p