The study protocol was approved by the Institutional Review Board of Asan Medical Center (2020-0271). The requirement for informed consent from individual patients was waived due to the retrospective design of the present study.

We retrieved consecutive 269 patients who underwent cardiac surgery because of IE between November 2013 and March 2019 from the institutional cardiac surgery database of Asan Medical Center, Seoul, Korea. The disease was considered active if the patients were currently being treated with anti-microbial agents against endocarditis based on the definition provided by the Society of Thoracic Surgeons (STS).5) Surgical treatment was considered in patients with following signs referred to the current practice guidelines; heart failure, severe valve dysfunction, periannular involvement or paravalvular abscess, large mobile vegetation (>10 mm) with embolism or high risk of embolic event, persistent sepsis despite adequate antibiotics therapy and IE caused by Staphylococcus aureus, fungal or highly resistant microorganism.6)7) Of 269 patients, 56 had PVE and remainder (n=213) had NVE. For patients undergoing surgery because of infective endocarditis, valve substitute was generally selected as for any other valve surgeries based on the current practice guidelines.6)7) However, bioprosthetic valve was occasionally considered in younger patients who had high operative risk, particularly, including chronic kidney and liver disease to avoid anticoagulation therapy.

Clinical follow-up information on individual patients was obtained through March 2020 from institutional electronic medical databases and chart reviews. Infectious etiology was collected by microbiologic laboratory results, and preoperative blood cultures, intraoperative cultures or 16S rDNA analysis of valvular tissue were reviews to confirm the causative pathogens. Detailed lesions of the affected valves were recorded based on echocardiographic reports, intraoperative findings and pathologic examinations of valvular tissue. Data regarding vital status and date of death were collected from institutional electronic, and were further validated by health claims databases for National Health Insurance Service, which is mandatory universal health insurance program providing comprehensive medical care coverage to all residents of the Republic of Korea.8)9)

The primary outcomes of interests were all-cause mortality and valve-related adverse events, including valve reinfection, stroke, valve reoperation and permanent pacemaker (PPM) implantation referencing the Society of Thoracic Surgery Guidelines.5) Early mortality was defined as deaths within 30 days of surgery. Recurrent IE was further categorized as early and late recurrence based on a cutoff of 180 days from the surgery.10)11) For further measure, perioperative complications (e.g., low cardiac output syndrome [LCOS] requiring mechanical circulatory support, early stroke, newly initiated dialysis and bleeding demanding surgical intervention) were also evaluated.

Continuous variables, expressed as mean±standard deviation, were compared using Mann-Whitney U test and Student's t test, as appropriate. Categorical variables were presented as frequencies and percentages and compared using χ² or Fisher exact tests. Time trends in proportions of PVE among surgical patients with IE were tested by χ² test for linear trend after the breaking down the surgical period into quartiles. Kaplan-Meier plots were used to construct the conditional probabilities for survival and freedom from adverse events, and the log-rank test was adopted to compare differences in the estimates between the 2 groups. Potential risk factors for overall mortality were assessed using Cox proportional-hazard (PH) model. In this model, the PH assumption was examined using scaled Schoenfeld residuals and no relevant violations were observed in all tested models. After univariable Cox PH analyses, variables with p value less than 0.2 were included into the multivariable PH models. Afterward, stepwise backward elimination technique was used to retain only variables with p<0.1 in the final multivariable model. Results are presented as hazard ratio (HR) for overall mortality with corresponding 95% confidence interval (CI) and p value. In addition, we assumed that the comparative risk of mortality may differ between PVE and NVE across early and late phases. Therefore, we conducted a landmark survival analysis to test this hypothesis. All reported p values were 2-sided, and p values less than 0.05 were considered statistically significant. Analyses were performed using R statistical software, version 3.4.0 (R Foundation, Vienna, Austria; http://www.R-project.org/).

The proportion of PVE among the entire surgical IE has gradually increased from 15.4% (11/71) to 29.5% (18/61) through the study period (p=0.055) (Figure 1). The baseline profiles of patients with PVE and NVE are summarized in Table 1. PVE patients were older, and more commonly had mitral aortic intervalvular fibrosa (MAIVF) involvement and abscess formation compared to NVE patients. There were 51 (19.0%) patients who experienced preoperative embolic event including cerebrovascular accident and visceral embolism; 8 (14.4%) and 43(20.2%) in PVE and NVE group, respectively (p=0.417). The mitral valve was the most frequently affected site in overall cohort (n=130, 48.3%) with its higher proportion in NVE (n=111, 52.1%) than in PVE (n=19, 33.9%). Inversely, PVE patients more often had IE involvement in the aortic valve (n=25, 44.6%). The distribution of infected valves is demonstrated in Table 1.

Indication of surgery and operative profiles are detailed in Tables 2 and 3, respectively. Of 269 patients, 81 (30.1%) patients underwent emergent surgery and 2 (0.7%) demanded preoperative mechanical circulatory support with extracorporeal membrane oxygenation because of cardiogenic shock. Valves were replaced with bioprosthetic and mechanical valve in 88 (32.7%) and 123 (45.7%) patients, respectively. A significantly higher proportion of PVE patients necessitated MAIVF repair compared to NVE patients. Cardiopulmonary bypass (235.1±105.3 vs. 133.3±51.0 minutes; p<0.001) and aortic cross clamping times (145.2±66.6 vs. 92.9±14.4 minutes; p<0.001) were significantly longer in PVE patients as compared with NVE patients.

The most prevalent pathogen in overall cohort was the viridans group Streptococci, which was isolated in 87 (32.3%) patients. The coagulase negative Staphylococci and viridans group Streptococci was the most common pathogen for patients with PVE (n=15, 26.8%) and NVE (n=78, 36.6%), respectively. Both methicillin-sensitive (5.4% vs. 7.5%; p=0.790) and resistant (5.4% vs. 5.2%; p=1.000) Staphylococcus aureus infection as well as fungal infection (5.4% vs. 1.9%; p=0.325) rates were similar between the groups (Table 4).

Details of perioperative and long-term adverse clinical outcomes, and their comparative risks are depicted in Table 5. The early mortality rates were 14.3% and 6.1% in PVE and NVE patients, respectively (p=0.049). With regard to perioperative complications, an early stroke rate was comparable between the groups (12.5% vs. 8.5%; p=0.356), but the incidence rates of other early major morbidities including LCOS demanding mechanical support (12.5% vs. 3.8%; p=0.017), newly initiated dialysis (19.6% vs. 6.6%; p=0.004), reoperation for bleeding (14.3% vs. 3.3%; p=0.003) and early PPM implantation (12.5% vs. 1.4%; p=0.001) rates were significantly higher in the PVE group than the NVE group. PVE patients had longer intensive care unit stay (3.0 days [interquartile range; IQR, 2.0–7.3 days] vs. 2.0 days [IQR, 1.0–3.0 days]; p=0.031) and hospital stay (35.5 days [IQR, 22.8–78.0 days] vs. 28.0 days [IQR, 15.0–37.0 days]; p=0.004) compared with NVE patients.

During follow-up (28.2 months [IQR, 11.5–45.8 months]; overall 648.8 patient-years), PVE group showed significantly higher risk of overall mortality (HR, 2.97; 95% CI, 1.73–5.10; p<0.001), valve reinfection (HR, 16.14; 95% CI, 3.24–80.35; p<0.001) and PPM implantation than NVE group (HR, 8.35; 95% CI, 2.67–26.18; p<0.001). With regard to survival, PVE group showed significantly higher risk of perioperative (odds ratio [OR], 2.56; 95% CI, 0.97–6.44; p=0.049) and overall mortality (HR, 2.97; 95% CI, 1.73–5.10; p<0.001). The landmark survival analysis revealed that mortality risk in PVE group was considerably higher within 1 year after IE surgery (HR, 3.09; 95% CI, 1.65–5.78; p<0.001) and such trend continued even after 1 year of time (HR, 2.76; 95% CI, 0.96–7.99; p=0.061). Recurrence of IE was documented in 7 patients, occurring 5 (5.1%/patient-years) in PVE group and 2 (0.2%/patient-years) in NVE group. Of 7 patients with valve reinfection, 3 had non-viridans group Streptococcal infection at their initial endocarditis event. The mean time after surgery to valve reinfection was 23.9 months (IQR, 7.6–23.9 months) in the PVE group and 33.9 months (IQR, 25.6–42.2 months) in the NVE group. Early recurrent IE occurred in only 1 patient with PVE at 5 months after index IE surgery. Late recurrence developed in 6 including 4 in PVE group and 2 in NVE group. In addition to 10 patients who had implanted PPM during hospitalization, 3 patients necessitated PPM implantation during a follow-up period (1 in the PVE group and 2 in the NVE group). Overall, PPM implantation was implanted in 13 patients in study cohort—8 (8.1%/patient-years) in the PVE group and 5 (0.5%/patient-years) in the NVE. The most common reason for PPM implantation was complete atrioventricular block (8/13, 61.5%) with its rates among all PPM cases of 62.5% (5/8) in the PVE group and 60.0% (3/5) in the NVE group.

Kaplan-Meier curves showed a significantly decreased overall survival (OS), and freedom from valve reinfection and PPM implantation in the PVE compared with NVE; however, there were no significant differences in freedom from stoke and valve reoperation between PVE and NVE group (Figure 2). One-year and 3-year OS rates were 69.6±6.1% and 61.3±6.7% in PVE patients, respectively; and 89.2±2.1% and 85.4±2.5% in NVE patients (p<0.001), respectively (Figure 2A). The freedom from IE recurrence at 3-year was 80.9±7.4% and 99.3±0.7% in the PVE and NVE group, respectively.

The results of univariable and multivariable Cox PH analyses to determine risk factors for overall mortality are summarized in Table 6. Final multivariable model revealed that PVE (HR, 2.67; 95% CI, 14.0–5.07; p=0.003), age (HR, 1.04; 95% CI, 1.01–1.07; p=0.009), chronic kidney disease (HR, 2.93; 95% CI, 1.58–5.44; p=0.001), multi-valve involvement (HR, 2.13; 95% CI, 1.14–3.98; p=0.01), and causative pathogens of Staphylococcus aureus (HR, 3.26; 95% CI, 1.68–6.33; p<0.001) or fungus (HR,3.13; 95% CI,1.25–7.86; p=0.015) were independent and significant risk factors for overall mortality

With regard to IE recurrence, comprehensive risk factor analyses were unavailable because of a limited number of events (n=7). Univariable Cox PH model, however, showed that following variables were significantly associated factors of recurrence of IE: PVE (HR, 14.05; 95% CI, 2.72–72.68; p=0.002), MAIVF involvement (HR, 17.92; 95% CI, 3.35–95.81; p=0.001), and causative pathogen of Staphylococcus aureus (HR, 10.83; 95% CI, 1.63–72.23; p=0.014) and fungus (HR, 19.87; 95% CI, 1.59–248.05; p=0.020).