Outcomes of Prosthetic Valve Endocarditis Managed by a Multidisciplinary Endocarditis Team: A Retrospective Cohort Study

Abstract

Background

Prosthetic valve endocarditis (PVE) is a serious complication of valve replacement associated with significant morbidity and mortality. However, outcomes for medical vs surgical strategies remain incompletely characterized.

Methods

We conducted a retrospective cohort study of adult patients with PVE, defined by modified Duke criteria, evaluated by a multidisciplinary endocarditis team at a tertiary-care centre between September 2021 and February 2024. Demographic, clinical, management, and outcome data were collected. Logistic regression analyses were performed to identify factors associated with in-hospital mortality.

Results

Among 67 patients with PVE, 58.2% (n = 39) were managed medically, and 41.8% (n = 28) underwent surgical intervention during the index hospitalization. Baseline characteristics and comorbidities were similar across groups. The in-hospital mortality incidence was 7.7% in the medical cohort and 10.7% in the surgical cohort, and the incidence of 90-day mortality was 30.8% and 21.4%, respectively. Surgical patients had a longer median length of stay (28 days vs 15 days). Readmission and reinfection rates were comparable. Acute renal failure was the only independent predictor of in-hospital mortality on multivariable analysis (odds ratio, 9.61; 95% confidence interval, 1.19-77.67; P = 0.04). Medical management was not independently associated with increased in-hospital mortality (odds ratio, 0.97; 95% confidence interval, 0.14-6.68; P = 0.97).

Conclusions

In this cohort of patients with PVE, individualized treatment guided by a multidisciplinary team was associated with favourable short-term outcomes. Medical therapy may be a safe alternative in selected patients without surgical indications. Early identification of acute renal dysfunction may assist in prognostication and inform management decisions.

Prosthetic valve endocarditis (PVE) is a serious and potentially life-threatening complication following valve replacement surgery, accounting for up to 20% of all cases of infective endocarditis (IE).1, 2 Infection rates with prosthetic valves are estimated to be up to 10 times higher than rates with native valves, with an incidence ranging from 0.3% to 2% per patient-year.2, 3, 4 The risk of developing endocarditis is greatest within the first year after valve implantation.5, 6, 7 Among all forms of IE, PVE carries the highest morbidity and mortality burden, particularly in patients with early-onset disease.^1,3, 4, 5, 6, 7, 8 The in-hospital mortality incidence is typically reported to be between 20% and 30%, but it may reach as high as 59% in some series.^3,9,10 Among persons who inject drugs (PWIDs), the presence of a prosthetic valve has been associated with increased risk of recurrent and complicated IE.¹¹ In addition, many patients with prosthetic valves are older and have significant comorbidities, which contribute to worse outcomes compared with those of younger patients with native-valve IE.¹²

The diagnosis of PVE is often challenging and typically requires multimodal imaging, including transesophageal echocardiography (TEE), cardiac computed tomography (CT), and/or cardiac positron emission tomography (PET).^2,9,13 Once the diagnosis is established, clinicians must determine whether to pursue medical therapy alone or combined medical and surgical management. Current American and European guidelines recommend surgery for patients with hemodynamic instability, paravalvular abscesses, congestive heart failure, persistent bacteremia, recurrent embolic events, or fungal infections.^14,15 Several factors—including heart failure, infection with Staphylococcus aureus, persistent bacteremia, and the presence of an abscess—have been independently associated with increased mortality in patients with PVE.^2,12,16, 17, 18, 19

Although existing literature supports a survival benefit with surgical intervention, most evidence is retrospective and observational in nature. Approximately 50% of patients with PVE undergo reoperation, often influenced by the severity of disease, the presence or absence of definitive surgical indications, and concerns regarding operative risk.²⁰ The implementation of multidisciplinary endocarditis teams has been shown to reduce in-hospital mortality in patients with IE and is hypothesized to aid in optimizing surgical decision-making.^21,22 However, only limited data exist on clinical outcomes specifically among patients with PVE managed by a formal multidisciplinary team approach. In this study, we describe our experience managing PVE at a tertiary-care centre using a multidisciplinary endocarditis team over a 2.5-year period.

Patients and Methods

Study design and patient identification

In September 2021, University of Kentucky Healthcare (UK) established a multidisciplinary endocarditis team (MDET) and a dedicated cardiovascular infectious diseases consult service (CVIDCS) to enhance the management of patients with complex endovascular infections. The composition of the team and its clinical protocol have been described previously (Supplemental Table S1).23, 24, 25, 26 The MDET comprises providers from the disciplines of infectious diseases, cardiac surgery, cardiology, addiction medicine, neurosurgery, neurology, physical medicine and rehabilitation, palliative care, and ethics. The group developed an internal protocol for medical treatment and formally meets weekly to discuss all inpatients with IE and document its recommendations in the electronic medical record. Cases for the conference are identified primarily by the CVIDCS, in collaboration with the cardiac surgery service. Frequent communication between MDET providers and clinical primary teams also occurs outside of scheduled meeting times. The CVIDCS is an interdisciplinary team housed in the division of infectious diseases and comprises an attending physician, an advanced practice provider, a nurse navigator, a pharmacist, and a social worker. The CVIDCS sees all inpatients with IE and follows them throughout their hospitalization and transition to outpatient.²⁵ Concurrent with the launch of the MDET, a prospective registry was created to capture data on all patients presented during weekly multidisciplinary case conferences. Institutional review board approval was obtained to establish the registry and to permit retrospective collection of demographic, diagnostic, treatment, and outcome-related information.

Between September 2021 and February 2024, a total of 414 patients were entered into the registry. For the present analysis, we included patients with suspected PVE, independent of whether they met the modified Duke criteria for definite PVE, during their index hospitalization.²⁷ For patients who did not return for follow-up at our institution, the electronic medical record was reviewed to assess post-discharge outcomes, including 90-day, and when available, 1-year mortality, based on documentation from other medical facilities in the region. The level of follow-up was 100% for determining the 90-day mortality rate, either at the authors’ institution, or through review of encounters at other hospitals with shared electronic medical records. The reporting of this study conformed to the Strengthening the Reporting of Observational Studies in Epidemiology statement.²⁸

Definition of terms

Injection drug use was defined as active intravenous substance use within 30 days of admission. Hepatitis C infection was defined by the presence of detectable viral load via polymerase chain reaction testing on admission. Persistent bacteremia was defined by blood cultures that were positive for ≥ 72 hours from the initial positive culture. Heart failure was defined as new-onset or acute worsening of symptomatic heart failure as outlined by the New York Heart Association heart failure classification system during the index hospitalization. Acute renal failure (ARF) was defined as requiring renal replacement therapy during the index hospitalization. All-cause readmissions were admissions within 30-days of discharge after the index hospitalization for any diagnosis. Admission for reinfection was defined as admissions after the index hospitalization, with a new diagnosis of IE determined to be separate from the initial episode of PVE. Thirty-day and 90-day mortality were calculated from the date of discharge (day 0). Patients who were transitioned to inpatient hospice units were also considered discharged on the day of transition. The duration of intravenous antibiotic therapy was defined as starting on the day of blood culture clearance. Oral suppressive antibiotics were defined as antimicrobials that were started after completion of an initial intravenous antibiotic course to prevent infection relapse rather than to cure the index infection.

Statistical analysis

Categorical variables were compared using Fisher’s exact test, and continuous variables with the Mann–Whitney U test. All tests were 2-sided with a significance threshold of P < 0.05. We first performed univariable logistic regression to identify factors associated with in-hospital mortality. Variables with P < 0.20 in univariable analyses were considered for multivariable modelling. Because some predictors may occur after admission, we fitted 2 separate multivariable models, as follows: (i) a model limited to baseline variables available at or near admission; and (ii) a model including in-hospital exposures. Medical vs surgical management was prespecified as a clinically relevant variable and was retained in multivariable models, regardless of statistical significance. Given the limited number of in-hospital deaths (n = 6), we used Firth’s penalized likelihood logistic regression to mitigate small-sample bias and instability from sparse data.

Model discrimination and calibration were assessed. Discrimination was measured using the C-statistic (area under the curve [AUC]) with 200-sample bootstrap internal validation; optimism-corrected AUCs were obtained by subtracting the mean optimism value from the apparent AUC. Calibration was evaluated using intercept and slope, also corrected for optimism. Calibration plots displayed observed vs predicted risks by deciles, with locally estimated scatterplot smoothing (LOESS).

For time-to-event analyses, Kaplan–Meier methods were used to compare survival between the medical and surgical groups. The time origin was date of discharge (day 0); follow-up continued until death or last contact. Patients who died during the index hospitalization (n = 6) were excluded from the Kaplan–Meier analysis because they had no post-discharge risk time. Curves were compared with the log-rank test; Wilcoxon (Breslow) and likelihood-ratio tests were examined as secondary comparisons. Kaplan–Meier figures included numbers at risk and pointwise 95% confidence bands with the log-rank P-value reported in the caption. Analyses were conducted using SAS (SAS Institute, Cary, NC) and R (RStudio, Posit Software).

Results

Between September 7, 2021 and February 1, 2024, a total of 67 patients were identified as having suspected PVE and were included in the analysis (Fig. 1). Of these, 39 (58.2%) were managed with medical therapy alone, and 28 (41.8%) underwent surgical intervention during the index hospitalization. Baseline characteristics were broadly similar between groups (Table 1). The median age was 62 years (interquartile range [IQR], 46.5-73) in the medical group and 57 years (IQR, 35-69.3) in the surgical group (P = 0.20). The majority of patients were male (67.7%; n = 26 vs 71.4%; n = 20; P = 0.88), and White (94.9%; n = 37 vs 96.4%; n = 27; P = 0.32). Injection drug use was present in 23.1% (n = 9) of medically managed patients and 17.9% (n = 5) of surgical patients (P = 0.61). Microbiologic data were similar between the 2 groups (Supplemental Table S2). The prevalence of prior surgical valve replacement was high in both cohorts, with 76.9% (n = 30) of medical and 82.1% (n = 23) of surgical patients having prosthetic valves in place for more than 1 year (P = 0.61; Supplemental Table S3). Transthoracic echocardiography was performed in 82.1% (n = 32) of medical patients and 75.0% (n = 21) of surgical patients (P = 0.48), and transesophageal echocardiography was obtained more frequently in the surgical group (92.9%; n = 26 vs 38.5%; n = 15; P < 0.01). All patients underwent either transthoracic or transesophageal echocardiography as part of their evaluation for PVE. Aortic valve involvement was significantly more common in the surgical group (71.4%; n = 20 vs 41.0%; n = 16; P = 0.01), and paravalvular abscesses were identified in 25.0% (n = 7) of surgical patients and 12.8% (n = 5) of medical patients (P = 0.20). Intensive care unit admission was more common in the medical group (43.6%; n = 17 vs 28.6%; n = 8; P = 0.21), as were vasopressor use (23.1%; n = 9 vs 10.7%; n = 3; P = 0.20) and mechanical ventilation (33.3; n = 13% vs 3.6%; n = 1; P < 0.01). Cardiothoracic surgery consultation was obtained for all surgical patients and for 74.4% of medical patients ((n = 29; P < 0.01).

Figure 1.

Study enrollment flowchart for patients with prosthetic valve endocarditis managed by a multidisciplinary endocarditis team. IE, infective endocarditis.

Table 1.

Demographics and clinical variables of patients with prosthetic valve endocarditis managed by a multidisciplinary endocarditis team

Variable	Medical management (n = 39)	Surgical management (n = 28)	P
Age, y	62 (46.5–73)	57 (35–69.3)	0.20
Male sex	67.7 (26)	71.4 (20)	0.88
Female sex	33.3 (13)	28.6 (8)
White	94.9 (37)	96.4 (27)	0.32
Asian	0 (0)	3.6 (1)
African American	5.1 (2)	0 (0)
Injection drug use	23.1 (9)	17.9 (5)	0.61
Hepatitis C viremia	5.1 (2)	14.3 (4)	0.20
Dental disease	35.9 (14)	46.4 (13)	0.39
Diabetes	30.8 (12)	35.7 (10)	0.68
Chronic dialysis	12.8 (5)	3.6 (1)	0.20
Prior sternotomies	97.4 (38)	78.6 (22)	0.01
CIED present	33.3 (13)	7.1 (2)	0.01
Prior admissions for infectious endocarditis	46.2 (18)	46.4 (13)	1.0
Days from admission to outside hospital or UK to valve surgery	NA	15.5 (7.8–25.5)
Days from arrival to UK to valve surgery	NA	13.5 (4.8–20.5)
Definite Duke criteria	61.5 (24)	71.4 (20)	0.58
Possible Duke criteria	30.8 (12)	17.9 (5)
Rejected Duke criteria	7.7 (3)	10.7 (3)
Transthoracic echocardiogram	82.1 (32)	75.0 (21)	0.48
Transesophageal echocardiogram	38.5 (15)	92.9 (26)	< 0.01
Cardiac PET/CT	5.1 (2)	7.1 (2)	1.0
Outside hospital transfers	35.9 (14)	42.8 (12)	0.57
Days of admission at outside hospital	4.5 (1.3–5)	3 (1.8–6)	0.53
Infectious disease consults	100 (39)	100 (28)	1.0
Days to infectious disease consult	1 (1–2.5)	1 (1–2)	0.53
Cardiothoracic surgery consult	74.4 (29)	100 (28)	< 0.01
Days to cardiothoracic surgery consult	1 (0–3.5)	0 (0–1.3)	0.16
ICU Admission	43.6 (17)	28.6 (8)	0.21
ICU length of stay	9 (6–21)	11 (8.8 –18.5)	0.10
Vasopressor requirement	23.1 (9)	10.7 (3)	0.20
Ventilator requirement	33.3 (13)	3.6 (1)	< 0.01
Acute renal failure	5.1 (2)	17.9 (5)	0.09
Pittsburgh Bacteremia Score	0 (0–2)	0 (0–2)	0.43
Persistent bacteremia	20.5 (8)	17.9 (5)	1.0
Heart failure 2/2 to infectious endocarditis	30.8 (12)	46.4 (13)	0.20
Septic pulmonary emboli	7.7 (3)	0 (0)	0.14
Cerebral emboli	33.3 (13)	28.6 (8)	0.68
CIED extraction	2.6 (1)	7.1 (2)	0.38

Values are % (n), or median (interquartile range), unless otherwise indicated.

CIED, cardiac implantable electronic device; ICU, intensive care unit; PET-CT, positron emission tomography-computed tomography; NA, not applicable; UK, University of Kentucky.

Among the 28 patients who underwent surgery, aortic valve replacement was performed in 75.0% (n = 21), including aortic root reconstruction in 39.3% (n = 11; Table 2). Mitral and tricuspid valve replacements occurred in, respectively, 32.1% (n = 9) and 17.9% (n =5), and 21.4% (n = 6) underwent multivalve surgery. A second cardiac operation during the same admission was required in 17.9% (n = 5), and 10.7% of surgical patients (n = 3) required extracorporeal membrane oxygenation (ECMO) support. Intravenous antibiotic therapy was administered to 100% of medical and 89.3% (n = 25) of surgical patients (P = 0.39; Table 3), with median durations of 36.5 days (IQR, 13.5-41.3) and 38 days (IQR, 16-42), respectively (P = 0.86). Oral step-down antibiotic therapy was used in 12.8% of medical patients (n = 5) and 25.0% of surgical patients ((n = 7; P = 0.20), with median durations of 17 days (IQR, 5-21) and 19 days (IQR, 11-24), respectively (P = 0.74). Oral suppressive therapy after completion of primary treatment was used more frequently in medically managed patients (33.3%; n = 13 vs 17.9%; n = 5; P = 0.16).

Table 2.

Overview of surgical management for patients with prosthetic valves admitted for endocarditis

Variable	N = 28
Total aortic reconstructions with aortic valve replacement	39.3 (11)
Total aortic valve replacement	75.0 (21)
Total mitral valve replacements	32.1 (9)
Total tricuspid valve replacements	17.9 (5)
Total pulmonic valve replacements	7.1 (2)
Multi-valve surgery performed	21.4 (6)
Second operation required during initial admission	17.9 (5)
ECMO initiation	10.7 (3)

Values are % (n).

ECMO, extracorporeal membrane oxygenation.

Table 3.

Antibiotic treatment for patients with prosthetic valves admitted for endocarditis, managed medically and surgically

Variable	Medical management N = 39	Surgical management N = 28	P
Intravenous antibiotic therapy	100 (39)	89.3 (25)∗	0.04
Duration of intravenous antibiotic therapy	36.5 (13.5–41.3)	38 (16–42)	0.86
Oral antibiotic therapy for course completion	12.8 (5)	25 (7)	0.20
Days of course completion oral antibiotic therapy	17 (5–21)	19 (11–24)	0.74
Oral antibiotic therapy suppressive therapy	33.3 (13)	17.9 (5)	0.16
Days of suppressive oral antibiotic therapy	175 (170)	228.5 (260.5)	0.58

Values are % (n), or median (interquartile range), unless otherwise indicated.

^∗This cohort included patients with fungal endocarditis, who were treated with oral fluconazole, and Bartonella endocarditis, who were treated with oral antibiotics.

The total length of hospitalization was significantly longer in the surgical group (median, 28 days [IQR, 17-40.8] vs 15 days [IQR, 7.5-21.5]; P = 0.02; Table 4). In-hospital mortality occurred in 7.7% of medical patients (n = 3) and 10.7% (n = 3) of surgical patients (P = 0.67). Thirty-day mortality was 25.6% (n = 10) in the medical group and 14.3% (n = 4) in the surgical group (P = 0.27), and 90-day mortality was 30.8% (n = 12) and 21.4% (n = 6), respectively (P = 0.40). Reinfection-related readmissions within 90 days occurred in 1 patient (2.6%) in the medical group and 2 patients (7.1%) in the surgical group (P = 0.38). Thirty-day all-cause readmission occurred in 12.8% of medical patients (n = 5) and 21.4% of surgical patients ((n = 6; P = 0.35), and 90-day readmission rates were 20.5% (n = 8) and 32.1% (n = 9), respectively (P = 0.29). Among medically managed patients who survived beyond 90 days, 17.2% (n = 5) ultimately underwent valve replacement during a subsequent admission.

Table 4.

Inpatient outcomes of patients with prosthetic valves admitted for endocarditis, managed medically and surgically

Variable	Medical management N = 39	Surgical management N = 28	P
Total length of stay, d	15 (7.5–21.5)	28 (17–40.8)	0.02
Patient-directed discharge	2.6 (1)	7.1 (2)	0.38
Inpatient mortality	7.7 (3)	10.7 (3)	0.67
30-d mortality	25.6 (10)	14.3 (4)	0.27
30-d reinfection readmission	2.6 (1)	3.6 (1)	1.0
30-d total readmissions	12.8 (5)	21.4 (6)	0.35
90-d reinfection readmission	2.6 (1)	7.1 (2)	0.38
90-d total readmission	20.5 (8)	32.1 (9)	0.29
90-d mortality	30.8 (12)	21.4 (6)	0.40

Values are median (interquartile range), or % (n), unless otherwise indicated.

In the medical arm, 3 patients were lost to follow-up at 1 year. Among the remaining 36 patients, 16 deaths had occurred at 1 year (44.4%). In the surgical arm, 5 patients were lost to follow-up at 1 year. Among the remaining 23 patients, 7 deaths had occurred at 1 year (30.4%).

Follow-up care after discharge varied between groups. Surgical patients were more likely to follow up with cardiothoracic surgery (78.6%; n = 22 vs 15.4%; n = 6; P < 0.01; Supplemental Table S4); follow-up care with infectious diseases (57.1%; n = 16 vs 61.5%; n = 24; P = 0.72) and cardiology (46.4%; n = 13 vs 35.9; n = 14; P = 0.39) was similar between cohorts. No follow-up care at our institution was documented in 30.8% of medically managed patients (n = 12) and 17.9% of surgically managed patients ((n = 5; P = 0.24).

Table 5 (top section) summarizes associations between baseline patient characteristics and in-hospital mortality; no baseline predictors reached statistical significance. Table 5 (bottom portion) also presents the model using in-hospital exposures. The final Firth logistic regression included 3 prespecified predictors—medical vs surgical management, ARF, and intensive care unit stay—of which only ARF remained statistically significant (odds ratio, 9.61; P = 0.04). Internal validation with 200 bootstrap resamples showed good performance. The optimism-corrected C-statistic (AUC) was 0.814, compared with an apparent AUC of 0.863 (Supplemental Fig. S1). Calibration was acceptable. The apparent calibration slope was 1.16, and the intercept was 0.04. After optimism correction, the slope was 1.00, and the intercept was 0.00, indicating no systematic miscalibration. The calibration plot (Supplemental Fig. S2) showed close agreement between observed and predicted mortality across risk deciles, visualized with locally estimated scatterplot smoothing (LOESS).

Table 5.

Univariate and multivariable Firth logistic regression for in-hospital mortality, with baseline predictors at admission and in-hospital exposures

Baseline predictors at admission	Univariate analysis		Multivariate analysis
Variable	Odds Ratio (95% CI)	P	Odds Ratio (95% CI)	P
Age	1.01 (0.95–1.05)	0.55
Male compared to female	0.91 (0.15–5.38)	0.70
Duke criteria∗
Rejected endocarditis	Reference	NA
Possible endocarditis	NA	NA
Definite endocarditis	NA	NA
Race∗
Caucasian	Reference	NA
Black	NA	NA
Other race	NA	NA
Aortic valve involvement	NA	NA
Mitral valve involvement	NA	NA
Pulmonic valve involvement∗	NA	NA
Tricuspid valve involvement	1.833 (0.18–18.41)	0.87
Previous infective endocarditis diagnosis	0.53 (0.01–3.14)	0.91
Paravalvular abscess	2.55 (0.41–15.86)	0.32
Multivalvular involvement	1.16 (0.21–6.29)	0.86
Injection drug use	0.74 (0.08–6.89)	0.81
Chronic dialysis	2.24 (0.22–23.11)	0.26
Type 2 diabetes mellitus	2.21 (0.41–11.98)	0.79
Dental disease	0.27 (0.03–2.45)	0.34
Outside hospital transfer	3.08 (0.52–18.16)	0.35
Hepatitis C viremia∗	NA	NA
Septic pulmonary embolus	NA	NA
Cerebral embolus	1.11 (0.19–6.57)	0.91
In-hospital exposures
Acute renal failure therapy	14.25 (2.14–94.74)	0.006	9.61 (1.19 – 77.67)	0.04
Intensive care unit admission	10.25 (1.12–93.67)	0.039	5.92 (0.86 – 40.69)	0.07
Vasopressor requirement	0.91 (0.10–8.57)	0.93
Mechanical ventilator requirement	2.55 (0.41–15.86)	0.32
Acute heart failure secondary to infective endocarditis	0.89 (0.15–5.24)	0.89
Medical management compared to surgical management	0.69 (0.13–3.72)	0.67	0.97 (0.14-6.68)	0.97
Persistent bacteremia	2.27 (0.37–14.01)	0.38

Boldface indicates statistical significance.

CI, confidence interval; NA, not applicable.

^∗Logistic regression could not be performed for variables with events occurring in only one group.

Among patients discharged alive (n = 61), the median length of the follow-up period was 531 days (IQR, 64.5-703) in the medical management group, and 558 days (IQR, 131-784) in the surgical group. Survival incidence appeared to be higher after surgical management (n = 25) than after medical management (n = 36); however, the difference did not reach statistical significance (log-rank P = 0.084; Fig. 2). The separation between curves is most pronounced early after discharge and narrows with longer follow-up periods.

Figure 2.

Kaplan–Meier survival post-discharge by management strategy (medical vs surgical); in-hospital deaths excluded. Shaded bands = 95% confidence intervals. surv_days, days of survival.

Discussion

In this retrospective single-centre cohort of 67 patients with PVE managed by a multidisciplinary team, we compared outcomes for medical vs surgical treatment during the index hospitalization. Overall, 58.2% received medical therapy (n = 39), and 41.8% (n = 28) underwent surgery. The incidence of in-hospital mortality was low in both groups—7.7% (n = 3) with medical treatment and 10.7% (n = 3) with surgery. The 90-day mortality rate was 30.8% (n = 12) with medical treatment and 21.4% (n = 6) with surgery. The median length of stay was longer for surgical patients, whereas readmission and reinfection rates were similar between groups. In multivariable Firth logistic regression, ARF was the only independent predictor of in-hospital mortality. Treatment strategy, either medical or surgical, was not independently associated with in-hospital mortality.

PVE remains a highly morbid and challenging infection, with historical in-hospital mortality rates ranging from 20% to 30%, and higher mortality exceeding 50% in certain cohorts.^4,7,9,13 Risk factors consistently associated with poor outcomes in PVE include heart failure, persistent bacteremia, infection with Staphylococcus aureus, the presence of peri-annular complications such as abscesses, and central nervous system involvement.¹⁶ Surgical intervention, when feasible, is often advocated for patients meeting established criteria, and prior observational studies suggest that early surgery may reduce mortality.⁹ Despite these recommendations, the evidence base is predominantly retrospective, with treatment decisions heavily influenced by individual patient factors and institutional practices.

An interesting point to note is that among the numerous clinical variables assessed, ARF emerged as the only independent predictor of in-hospital mortality in both univariate and multivariate analyses. Although ARF has been identified as a poor prognostic factor in broader populations with IE, few studies have specifically characterized its impact among patients with PVE. The need for hemodialysis likely reflects both the severity of the systemic illness and multiorgan dysfunction, including septic shock and hemodynamic instability. In the context of PVE, ARF may also complicate antimicrobial management, increase susceptibility to other infections, and limit surgical candidacy, further exacerbating risk. Future studies should aim to further delineate the temporal relationship between ARF development and outcomes in PVE and evaluate whether early renal-protective strategies could improve prognosis.

A substantial increase in mortality occurred between discharge and 90-day follow-up in both groups. The majority of the deaths after discharge occurred in patients who had goals-of-care discussions during the index hospitalization and were discharged with hospice services. In the medical arm, 7 patients were discharged to hospice and ultimately passed away before 90 days, accounting for more than two-thirds of the deaths between discharge and 90 days. The other 2 patients died from complications of heart failure. In the surgical arm, 1 patient was discharged on hospice. In the other 2 patients who passed away after discharge, one died from complications of heart failure, and the cause of death could not be ascertained in the other. Similarly, much of the lack of follow-up in both groups was attributed to patients transitioning to hospice care.

In 2 of the 5 patients with perivalvular complications who were medically managed, the presence of a perivalvular abscess was unclear. Due to this uncertainty and the fact that the patients were felt to have a prohibitively high risk for surgery, owing to their older age and comorbidities, redo surgery was not offered. One patient was being evaluated for surgery but suffered a cardiac arrest before an operation could be scheduled. One patient was felt to have a very high operative risk and was managed medically for > 1 year before undergoing surgery at another institution after receiving a second surgical opinion. The final patient had a confirmed mitral annular abscess but was critically ill on continuous renal replacement therapy and with acute limb ischemia. This patient was not offered surgery due to these issues.

Our cohort's lower-than-expected mortality rate may be partially explained by several factors. First, the majority of patients had late-onset PVE, which is generally associated with better outcomes, compared to those of early-onset cases.^1,3, 4, 5, 6, 7, 8 Nearly 80% of our patients developed PVE more than 1 year after prosthesis implantation. Additionally, despite significant comorbidity burdens, careful patient selection for surgery and comprehensive medical management strategies likely contributed to an improved survival rate. The high degree of infectious diseases and cardiothoracic surgery collaboration may have facilitated more timely interventions, antibiotic stewardship, and imaging assessments, thereby preventing clinical deterioration.

Approximately 42% of patients underwent valve surgery during the index hospitalization, and an additional 13% underwent delayed surgical intervention, aligning with prior reports that up to 50% of PVE patients ultimately require surgery.²⁹ Surgical patients in our cohort had higher rates of definitive Duke criteria fulfillment, paravalvular complications, and aortic valve involvement, consistent with classical surgical indications. Notably, surgical intervention was associated with longer hospitalizations but did not confer an obvious early survival advantage in our cohort, reflecting the complexity of balancing operative risk against disease severity. Additional research could explore whether optimized perioperative management or novel surgical techniques could mitigate these risks and further improve outcomes.

Generally speaking, at our institution, patients with drug-use-related PVE will be offered surgery if an urgent, class-I indication is present (new-onset heart failure, paravalvular abscess, etc.), unless they are felt to be too ill to survive a redo operation (similar to how a patient without drug use would be managed). For patients with drug-use-related PVE without class-I surgical indications, our group often favours an initial strategy of medical management and treatment of the underlying substance-use disorder, with outpatient follow-up to discuss redo surgery once the substance-use disorder is in remission.

Rates of relapsed infection were also relatively low at 90 days. Over one-quarter of patients (26.8%) received suppressive antibiotics, including one-third of medically managed patients. Notably, 17% of surgically managed patients also received suppressive antibiotics. Only limited data are available to guide clinicians with respect to the role of suppressive antimicrobials; this area is in great need of research. The literature that does exist is primarily in the form of small retrospective case series, and current guidelines do not provide a recommendation on the role of suppressive therapy in management of PVE.^14,15,30, 31, 32 The available studies suggest that suppressive therapy is generally well tolerated, and given the high mortality rate associated with medically managed PVE, it may be a reasonable intervention in patients carefully selected by a multidisciplinary endocarditis team.^30,31 At present, providers and patients must make educated guesses as they weigh the risks of long-term antibiotic therapy with the potential of a possible catastrophic recurrence of PVE.

Multidisciplinary management likely played a pivotal role in optimizing outcomes. Prior studies have demonstrated that dedicated endocarditis teams can significantly reduce mortality and improve guideline-adherent care in IE. Our experience supports these findings, highlighting the importance of early cardiothoracic surgery consultation, infectious diseases input, advanced imaging utilization, and individualized decision-making. Future prospective studies could better define which specific elements of multidisciplinary management contribute most to improved survival, and whether standardized MDET protocols could be disseminated to other centres or if virtual MDET models of care can improve outcomes in resource-limited settings.

Limitations

Several limitations warrant consideration. The retrospective design and single-centre setting limit the study’s generalizability. Selection bias cannot be excluded, as patients referred to our centre may differ systematically from those managed elsewhere. Additionally, 1-year mortality data were incomplete for patients discharged within the past year, although a 90-day follow-up period was robust. The relatively low prevalence of early-onset PVE in our cohort may have favourably influenced overall mortality rates. In addition, due to the relatively small sample size, we were unable to include some clinical variables, particularly categorical variables with more than 2 categories, in the multivariable logistic regression analysis. Some categorical variables had very few events (often only 0-2 cases per group), which would have led to unstable estimates, wide confidence intervals, and potential model overfitting. As a result, our ability to identify additional independent predictors of mortality may have been limited. Future multicentre studies with longer follow-up periods are necessary to confirm the durability of our findings and further characterize the long-term impact of multidisciplinary care on survival and quality of life.

Conclusion

Our study demonstrates that with multidisciplinary evaluation and individualized management, both surgical and medical strategies for PVE can yield favourable short-term outcomes. ARF requiring dialysis emerged as the strongest predictor of in-hospital mortality, emphasizing the need for early recognition and management of renal dysfunction in this population. Our findings reinforce the critical role of multidisciplinary collaboration in navigating the complex therapeutic decisions inherent to PVE, and they suggest that in selected patients, medical management can be pursued safely without compromising outcomes. We acknowledge that our study's relatively small sample size and variable sparsity may limit the stability of regression modelling. Future efforts should focus on prospective validation of multidisciplinary team approaches, early identification of high-risk features such as ARF, and the development of targeted strategies to further reduce mortality in this vulnerable population. Multicentre collaborations also may help strengthen these findings and facilitate their broader applicability.