Propensity Score Matched Analysis of Mechanical Vs. Bioprosthetic Valve Replacement in Patients with Previous Stroke

Circ J 2018; 82: 2041 – 2048 ORIGINAL ARTICLE doi: 10.1253/circj.CJ-18-0003 Cardiovascular Surgery

Propensity Score Matched Analysis of Mechanical vs. Bioprosthetic Valve Replacement in Patients With Previous Stroke

Shao-Wei Chen, MD; Victor Chien-Chia Wu, MD; Yu-Sheng Lin, MD; Ching-Chang Chen, MD; Dong-Yi Chen, MD; Chih-Hsiang Chang, MD; Pao-Hsien Chu, MD; Pei-Chi Ting, MD; An-Hsun Chou, MD, PhD; Tien-Hsing Chen, MD

Background: This study compared the long-term outcomes of prosthetic heart valve replacement with mechanical or bioprosthetic valves in patients with prior stroke.

Methods and Results: In total, 1,984 patients with previous stroke who had received valve replacement between 2000 and 2011 were identified using the Taiwan National Health Insurance Research Database. Propensity score matching analysis was used. Ultimately, 547 patients were extracted from each group and were eligible for analysis. On survival analysis, the risks of all-cause mortality and recurrence of stroke were similar. The incidence of major bleeding was greater in the mechanical valve group than in the bioprosthetic valve group (P=0.040), whereas no difference was observed in re-do valve surgery. On subgroup analysis, the bioprosthetic valve was favored for older age (≥60 years) and previous gastrointestinal (GI) bleeding patients. The mechanical valve, however, was favored for younger patients (<60 years).

Conclusions: In patients with previous stroke, bioprosthetic valves had a lower incidence of complications connected to major bleeding than did the mechanical valves. Survival and stroke recurrence rates, however, did not differ between the 2 groups. We recommend bioprosthetic valves for patients >60 years or who have a history of GI bleeding.

Key Words: Long-term prognosis; Old stroke; Valve replacement

ioprosthetic valves have been increasingly used in age of the patient at the time of the valve replacement.2,3 valve replacement surgery in the recent 20 years. In The Veterans Affairs randomized trial, however, deter- B a recent paper, however, titled “Mechanical or bio- mined that 50% of deaths were not related to prosthetic logic prostheses for aortic-valve and mitral-valve replace- valve complications and emphasized that comorbidity was ment” in New England Journal of Medicine, Goldstone et key in determining patient outcomes.4 Furthermore, the al, using data from an administration database in California, indications for valves in patients with a specific comorbid- reported that mechanical prosthetic valve replacement was ity are still controversial, similar to the previous debate associated with significantly better survival than biopros- concerning dialysis patients.5 thetic valve replacement.1 This landmark study raised the Patients who have previously had a stroke when under- concern that there is in fact limited evidence supporting the going valve surgery are not uncommon. Cardiac surgery in implantation of bioprosthetic valves in diseased or elderly patients with previous stroke is associated with increased patients. The choice of prosthetic heart valve during valve mortality and an increased risk of postoperative stroke.6–9 replacement surgery is between the bioprosthetic valve A history of stroke may affect long-term outcome after and the mechanical valve for the majority of patients. valve replacement, and influences the decision making for Determining which to use is primarily dependent on the the type of valve selection. Few studies, however, have

Received January 5, 2018; revised manuscript received March 23, 2018; accepted April 11, 2018; released online May 24, 2018 Time for primary review: 31 days Division of Thoracic and Cardiovascular Surgery, Department of Surgery (S.-W.C.), Department of Neurosurgery (C.-C.C.), Department of Anesthesiology (P.-C.T., A.-H.C.), Chang Gung Memorial Hospital, Linkou Medical Center, Chang Gung University, Taoyuan City; Graduate Institute of Clinical Medical Sciences, College of Medicine, Chang Gung University, Taoyuan City (S.-W.C., C.-H.C.); Department of Cardiology (V.C.-C.W., D.-Y.C., P.-H.C.), Kidney Research Center, Department of Nephrology (C.-H.C.), Chang Gung Memorial Hospital, Linkou Medical Center, Taoyuan City; Department of Cardiology, Chang Gung Memorial Hospital, Chiayi Branch, Chiayi City (Y.-S.L.); and Department of Cardiology, Chang Gung Memorial Hospital, Keelung Branch, Keelung (T.-H.C.), Taiwan Mailing address: Tien-Hsing Chen, MD, Department of Cardiology, Chang Gung Memorial Hospital, Keelung, No. 5 Fuxing Street, Guishan District, Taoyuan City 33305, Taiwan. E-mail: [email protected] ISSN-1346-9843 All rights are reserved to the Japanese Circulation Society. For permissions, please e-mail: [email protected]

Circulation Journal Vol.82, August 2018 2042 CHEN SW et al.

Figure 1. (A) Subject enrollment between 2000 and 2011, extracted from the Taiwan National Health Insurance Research Database; and (B) no. bioprosthetic valve and mechanical valve replacements.

reported the long-term outcomes of cardiac surgery all medical services for all Taiwanese people. The NHIRD according to valve replacement type in patients with previ- is derived from the NHI program; thus, NHI claims data ous stroke. Furthermore, valve selection in this high-risk accurately represent the Taiwan population, with only patient group has not been previously addressed. Evaluating minimal data omitted and a small participation bias pres- the balance of risk and benefits between bleeding because ent. The claim data for each patient can be linked using a of anticoagulant therapy and prosthetic valve structure consistent data encryption process, which allows for con- deterioration is complex and multifactorial in this specific tinual tracking of all of the claims of any individual within population. Therefore, the aim of this nationwide popula- the NHI program. All personal information in the NHIRD tion-based cohort study was to compare mortality, recur- is anonymized and de-identified to protect privacy; thus, rence of stroke, major bleeding, and re-do valve surgery this study was exempt from a full review by the Ethics between bioprosthetic and mechanical valves using pro- Institutional Review Board of Chang Gung Memorial pensity score matching analysis. Hospital.

Subjects Methods We examined the NHIRD inpatient records of all patients Data Source admitted for valve replacement between January 2000 and The data were obtained from the Taiwan National Health December 2011. Both Taiwan NHI procedure codes Insurance Research Database (NHIRD). Taiwan has a (68016, 68017, and 68018), which are used for claims for mandatory government-operated universal health insur- reimbursement, and International Classification of Diseases, ance program, the National Health Insurance (NHI) pro- Ninth Revision, Clinical Modification (ICD-9-CM) program, which was launched in 1995 and covers almost all cedure codes (35.21, 35.22, 35.23, 35.24), were used to (99%) of the approximately 23 million residents of Taiwan extract data for patients who received a valve replacement as of December 2015,10 providing comprehensive medical (Figure 1A). In total, 21,547 patients who underwent valve care coverage and reimbursing the medical expenditure of replacement in any position during the study period in

Circulation Journal Vol.82, August 2018 Valve Replacement in Old Stroke 2043

Table 1. Clinical and Surgical Subject Characteristics vs. PSM Status Before PSM After PSM Variable Bioprosthetic Mechanical Bioprosthetic Mechanical Total P-value P-value valve valve valve valve n 1,984 606 1,378 – 547 547 – Age (years) 63.7±13.2 67.4±11.6 62.0±13.6 <0.001 66.7±11.7 66.3±12.6 0.532 Male gender 1,120 (56.5) 349 (57.6) 771 (56.0) 0.497 316 (57.8) 309 (56.5) 0.669 Stroke type 0.645 0.633 Ischemic 1,729 (87.1) 534 (88.1) 1,195 (86.7) 481 (87.9) 476 (87.0) Hemorrhagic 185 (9.3) 51 (8.4) 134 (9.7) 48 (8.8) 56 (10.2) Both 70 (3.5) 21 (3.5) 49 (3.6) 18 (3.3) 15 (2.7) Time from last stroke (years) 2.7±3.0 3.3±3.4 2.4±2.9 <0.001 3.1±3.3 3.0±3.3 0.606 Valve disease etiology 0.012 0.743 Rheumatic heart 667 (33.6) 180 (29.7) 487 (35.3) 164 (30.0) 167 (30.5) Infective endocarditis 335 (16.9) 96 (15.8) 239 (17.3) 85 (15.5) 93 (17.0) Degeneration and others 982 (49.5) 330 (54.5) 652 (47.3) 298 (54.5) 287 (52.5) Comorbidity Diabetes mellitus 545 (27.5) 178 (29.4) 367 (26.6) 0.208 158 (28.9) 160 (29.3) 0.894 Hypertension 1,208 (60.9) 422 (69.6) 786 (57.0) <0.001 368 (67.3) 370 (67.6) 0.897 CHF 898 (45.3) 302 (49.8) 596 (43.3) 0.007 272 (49.7) 264 (48.3) 0.629 CAD 1,069 (53.9) 357 (58.9) 712 (51.7) 0.003 324 (59.2) 311 (56.9) 0.426 PAD 129 (6.5) 46 (7.6) 83 (6.0) 0.192 43 (7.9) 41 (7.5) 0.820 Atrial fibrillation 951 (47.9) 314 (51.8) 637 (46.2) 0.022 280 (51.2) 267 (48.8) 0.432 Dialysis 86 (4.3) 34 (5.6) 52 (3.8) 0.064 30 (5.5) 29 (5.3) 0.894 COPD 246 (12.4) 86 (14.2) 160 (11.6) 0.108 79 (14.4) 75 (13.7) 0.728 Liver cirrhosis 54 (2.7) 20 (3.3) 34 (2.5) 0.294 17 (3.1) 18 (3.3) 0.864 GI bleeding history 494 (24.9) 178 (29.4) 316 (22.9) 0.002 151 (27.6) 162 (29.6) 0.462 Charlson score 3.4±2.1 3.5±2.0 3.3±2.1 0.037 3.5±2.0 3.5±2.2 0.965 Hospital level <0.001 0.102 Medical center (teaching 1,336 (67.3) 373 (61.6) 963 (69.9) 361 (66.0) 335 (61.2) hospital) Regional/district hospital 648 (32.7) 233 (38.4) 415 (30.1) 186 (34.0) 212 (38.8) Valve type 0.054 0.744 AVR alone 772 (38.9) 260 (42.9) 512 (37.2) 237 (43.3) 225 (41.1) MVR alone 984 (49.6) 281 (46.4) 703 (51.0) 251 (45.9) 263 (48.1) AVR+MVR 228 (11.5) 65 (10.7) 163 (11.8) 59 (10.8) 59 (10.8) Additional surgery CABG 481 (24.2) 164 (27.1) 317 (23.0) 0.052 143 (26.1) 138 (25.2) 0.729 Aortic surgery 111 (5.6) 38 (6.3) 73 (5.3) 0.385 34 (6.2) 32 (5.9) 0.800 TV repair or replacement 193 (9.7) 56 (9.2) 137 (9.9) 0.627 54 (9.9) 57 (10.4) 0.764 Data given as mean ± SD or n (%). AVR, aortic valve replacement; CABG, coronary artery bypass grafting; CAD, coronary artery disease; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; GI, gastrointestinal; MVR, mitral valve replacement; PAD, peripheral artery disease; PSM, propensity score matching; TV, tricuspid valve.

Taiwan were identified. Of these, 2,016 patients (9.36%) during the study period (Figure 1B). To minimize potential with a history of stroke were identified from ICD-9-CM selection bias, we calculated a propensity score from procedure codes 430.xx–437.xx. Patients were excluded if selected variables (Table 1) and matched each patient in they were aged <18 years, had missing demographic or the bioprosthetic valve group with each patient in the surgical information, or it was not their first admission for mechanical valve group. Finally, 547 patients from each valve replacement. In total, 1,984 adult patients with a his- group were identified and were eligible for analysis. tory of stroke were identified for first-time valve replacement. Comorbidity and Outcome Valve type was further identified using ICD-9-CM pro- Preoperative comorbidity and surgery details were identi- cedure codes (bioprosthetic valve codes 35.21 and 35.23 fied using ICD-9-CM and NHI procedure codes for the and mechanical valve codes 35.22 and 35.24). Patients were index hospitalization and prior hospitalizations (Table S1). more likely to undergo valve replacement with a mechani- We obtained inpatient records for ≥3 years before the cal valve (1,378 patients; 69.5%) than with a bioprosthetic index admission for all patients to track the history of valve (606 patients; 30.5%). The bioprosthetic valves, how- comorbidity. For example, we enrolled patients in 2000 ever, were increasingly used in valve replacement surgery and identified comorbidities from their medical records

Circulation Journal Vol.82, August 2018 2044 CHEN SW et al.

Table 2. Operation-Related Complications and Outcomes vs. PSM Status Before PSM After PSM Variable Total Bioprosthetic Mechanical Bioprosthetic Mechanical P-value† P-value‡ (n=1,984) value (n=606) valve (n=1,378) valve (n=547) valve (n=547) Categorical parameter In-hospital mortality 202 (10.2) 68 (11.2) 134 (9.7) 0.310 55 (10.1) 63 (11.5) 0.430 Any stroke 265 (13.4) 85 (14.0) 180 (13.1) 0.561 81 (14.8) 63 (11.5) 0.104 Ischemic stroke 231 (11.6) 72 (11.9) 159 (11.5) 0.826 68 (12.4) 57 (10.4) 0.293 Hemorrhagic stroke 39 (2.0) 15 (2.5) 24 (1.7) 0.281 15 (2.7) 8 (1.5) 0.139 Cardiogenic shock 210 (10.6) 75 (12.4) 135 (9.8) 0.086 63 (11.5) 61 (11.2) 0.855 requiring MCS Re-exploration for 57 (2.9) 16 (2.6) 41 (3.0) 0.681 15 (2.7) 17 (3.1) 0.715 bleeding Massive blood transfusion: 360 (18.1) 118 (19.5) 242 (17.6) 0.309 107 (19.6) 106 (19.4) 0.948 PRBC >10 U Hemodialysis (de novo 142 (7.2) 44 (7.3) 98 (7.1) 0.906 38 (6.9) 51 (9.3) 0.148 dialysis) Sepsis 123 (6.2) 49 (8.1) 74 (5.4) 0.022 44 (8.0) 30 (5.5) 0.093 Prolonged hospital stay 666 (33.6) 237 (39.1) 429 (31.1) 0.001 208 (38.0) 190 (34.7) 0.260 >28 days Continuous parameter ICU stay (days) 9.7±12.2 11.0±13.0 9.1±11.8 0.001 11.0±13.4 10.7±13.4 0.675 In-hospital cost (NTD×104) 52.6±25.3 55.6±27.0 51.3±24.5 <0.001 55.4±27.6 54.5±26.0 0.584 Data given as mean ± SD or n (%). †Unadjusted analysis (crude analysis); ‡adjusted for propensity score. Logistic regression for categorical parameters and linear regression for continuous parameters. ICU, intensive care unit; MCS, mechanical circulatory support; NTD, New Taiwan Dollar; PRBC, packed red blood cells; PSM, propensity score matching.

from 1997. The outcomes of interest in this study were all- Results cause mortality, recurrence of stroke, re-do valve surgery, and major bleeding. Death records and records of with- Subject Characteristics drawal from the NHI program were used to identify mor- Before matching, patients with bioprosthetic valve replace- tality, as validated in previous studies.11,12 Major bleeding ment were older and had a higher prevalence of comor- was defined as any admission after the index admission bidity (e.g., hypertension, heart failure, coronary artery with an ICD-9-CM diagnosis code 430*–432*, 578*, 719.1*, disease, atrial fibrillation, gastrointestinal (GI) bleeding 423.0, 599.7, 626.2, 626.6, 626.8, 627.0, 627.1, 786.3, 784.7, history, and higher Charlson score) than did those in the and 459.0 as the principal diagnosis. The date of admission mechanical valve group. The time between last stroke and for valve replacement was defined as the index hospitaliza- index admission, however, was longer in the bioprosthetic tion. The patients were followed from the index admission valve group (3.3 vs. 2.4 years). Regarding disease etiology to 31 December 2011 or the date of death, whichever came by valve, the bioprosthetic valve group had a lower inci- first. dence of rheumatic heart disease and infective endocarditis but a higher incidence of degeneration pathology than did Statistical Analysis the mechanical valve group. No differences, however, in We compared clinical and surgical characteristics between demographic and surgery details were observed between the 2 groups (bioprosthetic valve vs. mechanical valve) the 2 groups after propensity score matching (Table 1). using chi-squared test for categorical variables or t-test for continuous variables. Rates (categorical parameter) or In-Hospital Mortality and Postoperative Complications levels (continuous parameter) of postoperative complica- Table 2 lists the early results of valve replacement in tions between the 2 groups were compared on logistic patients with previous stroke according to valve type regression analysis and linear regression analysis, respec- before and after matching. The in-hospital mortality rate tively. Cumulative incidence of primary outcomes (i.e., of valve replacement surgery was 10.2%, and the incidence all-cause mortality) was compared between the groups, rate for postoperative stroke was 13.4%. Before propensity and log-rank test used to compare group differences. The score matching, a higher incidence of postoperative sepsis, risk factors of in-hospital mortality after valve replacement more prolonged in-hospital stay, longer intensive care unit in patients with previous stroke before matching were stay, and higher in-hospital medical expenditure was investigated using multivariable logistic regression analy- observed in the bioprosthetic valve group, but the inci- sis. Finally, the causes of major bleeding were compared dence of postoperative short-term outcomes was similar between the 2 groups using chi-squared test. Statistical between the 2 groups after propensity score matching. analysis was carried out using SPSS 22 (IBM SPSS, Armonk, NY, USA). Primary Outcomes During Follow-up Figure 2 illustrates the cumulative incidence of primary outcomes. The mortality rate was 40.6% and 40.0% (P=0.912, Figure 2A) and the incidence of recurrent stroke

Circulation Journal Vol.82, August 2018 Valve Replacement in Old Stroke 2045

Figure 2. Kaplan-Meier survival curves for (A) all-cause mortality, (B) any stroke, (C) re-do valve surgery, and (D) major bleeding vs. replacement valve type.

Figure 3. Cause of major bleeding vs. replacement valve type.

was 11.2% and 12.2% (P=0.680; Figure 2B) in the biopros- in terms of the proportion of intracerebral hemorrhage thetic valve and mechanical valve groups, respectively. and genitourinary bleeding (Figure 3). Likewise, the rate of re-do valve surgery was also similar between the 2 groups (Figure 2C). Risk of In-Hospital Mortality Before Matching The risk of major bleeding was higher in the mechanical The risk factors for in-hospital mortality in patients with valve group than in the bioprosthetic valve group (10.2% previous stroke before matching were advanced age; infec- vs. 6.8%, P=0.040, Figure 2D). The accumulated number tive endocarditis, degeneration, or other as valve disease of admissions because of major bleeding was 86 and 44 in etiology (compared with rheumatic heart); heart failure; the mechanical valve group and bioprosthetic valve group, dialysis; GI bleeding history; valve type of mitral valve respectively. With regard to major bleeding, the propor- replacement alone (compared with aortic valve replace- tions of GI bleeding and respiratory bleeding were greater ment alone), additional coronary artery bypass grafting in the mechanical valve group than the bioprosthetic valve (CABG) or aortic surgery (Table S2). group (P<0.05; Figure 3). No group difference was observed

Circulation Journal Vol.82, August 2018 2046 CHEN SW et al.

Figure 4. Hazard ratios for all-cause mortality vs. replacement valve type. AVR, aortic valve replacement; CABG, coronary artery bypass grafting; GI, gastrointestinal; MVR, mitral valve replacement.

All-Cause Mortality During Follow-up complications associated with major bleeding compared On subgroup analysis, the effect of valve type (biopros- with the mechanical valve, but survival, stroke recurrence, thetic vs. mechanical) was neutral on all-cause mortality in and re-do valve surgery did not differ between the 2 groups. terms of gender, stroke type, years to prior stroke, valve We further identified the age at valve replacement and that disease etiology, atrial fibrillation, valve type, and 3 addi- at which a history of previous GI bleeding might influence tional surgeries. This suggests that younger patients might the selection of valve type. benefit more from mechanical valve surgery than would Stroke is a manifestation of cardiovascular disease and older patients (P for interaction=0.037). In contrast, patients has a negative impact on outcome in surgical patients,6 with a history of GI bleeding might benefit more from potentially leading to long-term subsequent complications bioprosthetic valve surgery than would those without (P and mortality. Preoperative stroke or neurological dys- for interaction=0.047; Figure 4). function is a known risk factor for cardiac surgery and has been included in all current risk assessment models.13,14 Discussion Recently, Bottle et al demonstrated that preoperative stroke was associated with perioperative mortality dou- In this study, we compared the outcomes of mechanical bling, increasing the risk of postoperative stroke and pro- valve vs. bioprosthetic valve replacement in patients with longed hospital stay in 62,104 patients with CABG.7 previous stroke during a 12-year period using a national Limited information is available, however, on the long- database that covers nearly the entire population of term outcomes of valve replacement surgery in patients Taiwan. Finally, data on 547 patients from each group with previous stroke. Furthermore, no previous studies have were extracted and analyzed using propensity score match- focused on the impact of medical history of stroke on out- ing to ensure that the baseline patient characteristics were comes by comparing bioprosthetic and mechanical valves. similar. The in-hospital mortality rate was relatively high We next examined the history of valve disease guidelines for at 10.2%, and the incidence of postoperative stroke was patients with specific comorbidities. The American Heart 13.4%. To our knowledge, this is the first study to compare Association/American College of Cardiology (AHA/ACC) the long-term outcomes of valve replacement in this spe- valvular heart disease guidelines of 1998 previously fea- cific patient group. We found that the bioprosthetic valve tured a specific indication for valve selection in specific resulted in superior outcome with regard to reducing the populations, with class IIa recommended for mechanical

Circulation Journal Vol.82, August 2018 Valve Replacement in Old Stroke 2047 valves and class III recommended for bioprosthetic valves vant to valve type selection. Careful monitoring of INR in dialysis patients.15,16 The 2006 revised AHA/ACC valvu- after valve surgery and use of aggressive preventive agents lar heart disease guidelines, however, removed the previ- such as a proton pump inhibitors in patients who have ous statement on the basis of a large national database received mechanical valves are justified. study.5,17 Therefore, the present study provides information concerning valve selection in a specific group: patients Study Limitations with a history of stroke. Some limitations should be noted. First, the NHIRD does The choice of valve prosthesis type is based on the con- not include data on ventricular ejection fraction, neuro- sideration of several factors, including the age at valve logical status, or Society of Thoracic Surgeon (STS) score. implantation, the potential need for long-term anticoagu- The lack of these data may have influenced the analysis lation or reoperation, and, most vitally, patient preference. and the reported outcomes. Furthermore, the NHIRD The balance between the risk of reoperation for structural held no data regarding other prosthetic valve-related valve degeneration in bioprosthetic valves and the risk parameters such as valve size and brand. We, however, associated with long-term anticoagulant therapy should be adopted alternative variables as proxies for unavailable discussed with the patient. The ACC/AHA 2014 guidelines information, such as congestive heart failure, which is for valvular heart disease recommend using mechanical associated with left ventricular ejection fraction. Moreover, valves for patients aged <60 years, and bioprosthetic valves the accuracy of the NHIRD NHI procedure codes and for patients >70 years. Using either bioprosthetic or mechan- ICD-9-CM has been validated in patients with major car- ical valves is reasonable in patients between 60 and 70 years diovascular disease and cardiac surgery; the diagnosis code of age.3 On the basis of the present study, we recommend of stroke in NHIRD was also validated;12 the database a bioprosthetic valve for patients with previous stroke and appears to be a valid resource for population research in age >60 years. This corresponds to the current trend for cardiovascular research and is highly accurate in patients bioprosthetic valves in patients aged >60 years.18 Further with intervention.23 Second, the duration of follow-up investigation in the form of prospective randomized mul- might not have been long enough to demonstrate the dif- ticenter studies, however, is warranted. ference in re-do valve surgery because structural valve According to the present study, bioprosthetic valves are deterioration usually occurs >10 years after bioprosthetic associated with superior outcomes in reducing major valve implantation. Third, no data on anticoagulant or bleeding compared with mechanical valves. This corre- antiplatelet medication type or dosage during follow-up sponds to other studies that have directly compared bio- were available for use in this study; the influence of this prosthetic valve and mechanical valve use in the general lack of data on the identified outcomes of major bleeding population.18,19 No differences, however, in all-cause mor- should be acknowledged and evaluated. tality or recurrent stroke in patients with a history of stroke Finally, this study was conducted in patients based in were evident. The causes of major bleeding in patients with Asia. The incidence of rheumatic heart disease (30% in the a history of stroke are complex and include poor general present study) in south Asia is higher than in Western health, comorbidity and stress ulcer. We further identified countries and industrialized countries,22 and thus the that, of the major bleeding events, GI bleeding and respira- results may not be directly extrapolated to other countries. tory bleeding occur more often in the mechanical valve group Despite these limitations, we believe that the present study than in the tissue valve replacement group. Furthermore, provides a valuable contribution to the outcome analysis on subgroup analysis, a history of GI bleeding was associ- of valve replacement in patients with previous stroke. ated with superior long-term outcome in the bioprosthetic valve group relative to the mechanical valve group. Conclusions GI bleeding after cardiac surgery is a common complica- tion and is associated with high mortality and morbidity. Compared with the mechanical valve replacement group, Predictive factors must be identified early and preventive the bioprosthetic valve replacement group had a lower inci- measures must be applied to ensure the best possible out- dence of complications related to major bleeding. Survival comes.20 Mori et al noted an ethnicity effect in bleeding and recurrence of stroke during follow-up, however, did after valve surgery and recommended lower international not differ between these 2 groups. normalized ratio (INR) of prothrombin time after pros- In conclusion, for patients with previous stroke, we rec- thetic heart valve replacement surgery to prevent hemor- ommend the bioprosthetic valve for patients >60 years old rhagic complications in Asian patients.21 Krawiec et al or who have a history of previous GI bleeding. The present reported that instances of GI bleeding after cardiac surgery data provide further information on valve selection in are increasing and the most common cause is peptic ulcer, patients with specific comorbidity. The final decision-making which is related to Helicobacter pylori (H. pylori) infec- process, however, should still take into account patient tion.22 They recommended identification of this risk factor values and preferences in a detailed discussion. during preoperative evaluation alongside a test for H. pylori to decrease the risk of GI bleeding. Based on the Acknowledgments present findings, preoperative GI bleeding history critically affected in-hospital mortality after valve replacement sur- The study was based on data from the NHIRD provided by the National Health Insurance Administration of the Ministry of Health gery in patients with previous stroke. Additionally, preop- and Welfare, Taiwan, and managed by the National Health Research erative GI bleeding may have influenced valve type selection. Institutes of Taiwan. The interpretation and conclusions contained, Routine preoperative risk evaluation of GI bleeding however, do not represent those of the National Health Insurance including tests for occult blood in stool and H. pylori infec- Administration, Ministry of Health and Welfare, or the National Health Research Institutes of Taiwan. S.-W.C. and T.-H.C. had full tion should be performed. Additionally, use of a GI tract access to all the data used in the study and take responsibility for the endoscope in high-risk patients is reasonable and poten- integrity of the data and the accuracy of analysis. The authors thank tially beneficial because it may provide information rele- Alfred Hsing-Fen Lin for his assistance with statistical analysis.

Circulation Journal Vol.82, August 2018 2048 CHEN SW et al.

Disclosures surgery risk models: Part 2 – isolated valve surgery. Ann Thorac Surg 2009; 88: S23 – S42. The authors declare no conflicts of interest. 15. ACC/AHA guidelines for the management of patients with valvular heart disease. A report of the American College of Cardiology/American Heart Association Task Force on Practice References Guidelines (Committee on Management of Patients with Valvular 1. Goldstone AB, Chiu P, Baiocchi M, Lingala B, Patrick WL, Heart Disease). J Am Coll Cardiol 1998; 32: 1486 – 1588. Fischbein MP, et al. Mechanical or biologic prostheses for aor- 16. Bianchi G, Solinas M, Bevilacqua S, Glauber M. Are biopros- tic-valve and mitral-valve replacement. N Engl J Med 2017; 377: theses associated with better outcome than mechanical valves in 1847 – 1857. patients with chronic kidney disease requiring dialysis who 2. Nishimura RA, Otto CM, Bonow RO, Carabello BA, Erwin JP undergo valve surgery? Interact Cardiovasc Thorac Surg 2012; 15: 3rd, Guyton RA, et al. 2014 AHA/ACC guideline for the man- 473 – 483. agement of patients with valvular heart disease: A report of the 17. Bonow RO, Carabello BA, Kanu C, de Leon AC Jr, Faxon DP, American College of Cardiology/American Heart Association Freed MD, et al. ACC/AHA 2006 guidelines for the manage- Task Force on Practice Guidelines. J Thorac Cardiovasc Surg ment of patients with valvular heart disease: A report of the 2014; 148: e1 – e132. American College of Cardiology/American Heart Association 3. Nishimura RA, Otto CM, Bonow RO, Carabello BA, Erwin JP Task Force on Practice Guidelines (writing committee to revise 3rd, Guyton RA, et al. 2014 AHA/ACC guideline for the man- the 1998 Guidelines for the Management of Patients With agement of patients with valvular heart disease: Executive sum- Valvular Heart Disease): Developed in collaboration with the mary: A report of the American College of Cardiology/American Society of Cardiovascular Anesthesiologists: Endorsed by the Heart Association Task Force on Practice Guidelines. J Am Coll Society for Cardiovascular Angiography and Interventions and Cardiol 2014; 63: 2438 – 2488. the Society of Thoracic Surgeons. Circulation 2006; 114: e84 – 4. Hammermeister K, Sethi GK, Henderson WG, Grover FL, e231. Oprian C, Rahimtoola SH. Outcomes 15 years after valve 18. Zhao DF, Seco M, Wu JJ, Edelman JB, Wilson MK, Vallely replacement with a mechanical versus a bioprosthetic valve: MP, et al. Mechanical versus bioprosthetic aortic valve replace- Final report of the Veterans Affairs randomized trial.J Am Coll ment in middle-aged adults: A systematic review and meta-anal- Cardiol 2000; 36: 1152 – 1158. ysis. Ann Thorac Surg 2016; 102: 315 – 327. 5. Herzog CA, Ma JZ, Collins AJ. Long-term survival of dialysis 19. Glaser N, Jackson V, Holzmann MJ, Franco-Cereceda A, patients in the United States with prosthetic heart valves: Should Sartipy U. Aortic valve replacement with mechanical vs. biological ACC/AHA practice guidelines on valve selection be modified? prostheses in patients aged 50 – 69 years. Eur Heart J 2016; 37: Circulation 2002; 105: 1336 – 1341. 2658 – 2667. 6. Liao CC, Chang PY, Yeh CC, Hu CJ, Wu CH, Chen TL. 20. Viana FF, Chen Y, Almeida AA, Baxter HD, Cochrane AD, Outcomes after surgery in patients with previous stroke. Br J Smith JA. Gastrointestinal complications after cardiac surgery: Surg 2014; 101: 1616 – 1622. 10-year experience of a single Australian centre. ANZ J Surg 7. Bottle A, Mozid A, Grocott HP, Walters MR, Lees KR, Aylin 2013; 83: 651 – 656. P, et al. Preoperative stroke and outcomes after coronary artery 21. Mori T, Asano M, Ohtake H, Bitoh A, Sekiguchi S, Matsuo Y, bypass graft surgery. Anesthesiology 2013; 118: 885 – 893. et al. Anticoagulant therapy after prosthetic valve replacement: 8. Bucerius J, Gummert JF, Borger MA, Walther T, Doll N, Optimal PT-INR in Japanese patients. Ann Thorac Cardiovasc Onnasch JF, et al. Stroke after cardiac surgery: A risk factor Surg 2002; 8: 83 – 87. analysis of 16,184 consecutive adult patients. Ann Thorac Surg 22. Krawiec F, Maitland A, Duan Q, Faris P, Belletrutti PJ, Kent 2003; 75: 472 – 478. WDT. Duodenal ulcers are a major cause of gastrointestinal 9. Massaro A, Messe SR, Acker MA, Kasner SE, Torres J, Fanning bleeding after cardiac surgery. J Thorac Cardiovasc Surg 2017; M, et al. Pathogenesis and risk factors for cerebral infarct after 154: 181 – 188. surgical aortic valve replacement. Stroke 2016; 47: 2130 – 2132. 23. Cheng CL, Lee CH, Chen PS, Li YH, Lin SJ, Yang YH. Validation 10. Hsing AW, Ioannidis JP. Nationwide population science: Lessons of acute myocardial infarction cases in the National Health from the Taiwan National Health Insurance Research Database. Insurance Research Database in Taiwan. J Epidemiol 2014; 24: JAMA Intern Med 2015; 175: 1527 – 1529. 500 – 507. 11. Wu CY, Chen YJ, Ho HJ, Hsu YC, Kuo KN, Wu MS, et al. Association between nucleoside analogues and risk of hepatitis B virus-related hepatocellular carcinoma recurrence following Supplementary Files liver resection. JAMA 2012; 308: 1906 – 1914. Supplementary File 1 12. Cheng CL, Chien HC, Lee CH, Lin SJ, Yang YH. Validity of Figure S1. Sensitivity test: Kaplan-Meier survival curves for in-hospital mortality data among patients with acute myocardial stroke without transient ischemic stroke. infarction or stroke in National Health Insurance Research Database in Taiwan. Int J Cardiol 2015; 201: 96 – 101. Table S1. ICD-9-CM codes 13. Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Table S2. Risk of in-hospital mortality after valve replacement in Goldstone AR, et al. EuroSCORE II. Eur J Cardiothorac Surg patients with previous stroke before PSM (n=1,984) 2012; 41: 734 – 744. 14. O’Brien SM, Shahian DM, Filardo G, Ferraris VA, Haan CK, Please find supplementary file(s); Rich JB, et al. The Society of Thoracic Surgeons 2008 cardiac http://dx.doi.org/10.1253/circj.CJ-18-0003

Circulation Journal Vol.82, August 2018