Epidemiology of Left Ventricular Assist Device Infections: Findings From a Large Nonregistry Cohort

Abstract

Background

Infection is a major complication during circulatory support with a left ventricular assist device (VAD). Changes in device characteristics and treatment practices in the last decade can affect the epidemiology of infection. The International Society for Heart and Lung Transplantation (ISHLT) has published recommendations on the prevention and management of VAD infections, but data to support these recommendations remain sparse.

Methods

We performed a retrospective review of 455 patients who underwent VAD placement from 2009 to 2015. Infection episodes were defined using ISHLT criteria and were also grouped as endovascular or local. Analysis included descriptive statistics.

Results

There were 174 patients (38.6%) with a VAD infection. Infection incidence was 36.9 cases per 100 person-years of VAD support. The driveline was the most common infection site (67.2%). Systemic inflammatory response syndrome (SIRS) criteria were not satisfied in 29.2% of patients with endovascular infections, and computed tomography (CT) examinations were normal in 37.7% of cases. Gram-positive bacteria caused 65.6% of infections in patients with an available culture. Antimicrobial suppression was used in 72.3% of patients who survived treatment. Median survival after infection was 35 months for patients with VAD-related infections versus 14 months for patients with VAD-specific infections.

Conclusions

VAD infections continue to be a major complication after implantation. Clinical criteria alone were not predictive of serious infections, and many patients with confirmed infection had normal CTs. Patients with VAD-specific infections had lower median survival than patients with VAD-related infections.

We describe the incidence and epidemiology of left ventricular assist device (VAD) infections in a large nonregistry cohort of patients. VAD infections continue to be a frequent complication and are a heterogeneous group of entities with variable influence in outcomes.

(See the Editorial Commentary by Pericàs on pages 198–201.)

It is estimated that over 7 million Americans have a current diagnosis of heart failure [1]. Patients with end-stage heart failure are poorly responsive to medical management, and mechanical circulatory support with ventricular assist devices (VADs) can be used as a bridge to heart transplantation or long-term destination treatment [2]. With rates of heart failure expected to grow to as high as 500 000 new cases per year by 2031, it is estimated that over 250 000 individuals will be candidates for VAD placement [3]. Despite significant improvement in cardiovascular parameters and quality of life, the long-term outcomes in VAD recipients are significantly limited by infectious complications [4]. VAD-related sepsis is the second most common cause of death in the 3 months post–VAD implantation, and the risk of infectious complications continues to be elevated after this period [5]. Although prior publications described the epidemiology of VAD infections, these studies were limited by small sample sizes (ranging from 68 to 247 VAD patients with only 36–78 patients with infection) [6–9]. Definitions for VAD infections were only standardized in 2011, so work prior to this date may also have inconsistent results based on the definitions used for infection [6, 10]. Changes in device characteristics, implantation, and other treatment practices [11–14] make it important to study the epidemiology of VAD infections using recent data. In addition, the International Society for Heart and Lung Transplantation (ISHLT) published recommendations on the prevention and management of VAD infections in 2017; however, supporting data are sparse. We performed a retrospective cohort study from 2009–2015 to describe the epidemiology of VAD infection in a large cohort of patients from a single center using the standardized ISHLT consensus definitions [15] as well as grouped by endovascular and local infections per prior studies to allow for comparisons with prior and recent works [9, 16]. The cohort was also used to assess the clinical applicability of some of the published ISHLT guidelines for prevention and treatment of VAD infections [17].

METHODS

This was a retrospective review of 455 consecutive patients who underwent VAD implantation from January 2009 to May 2015 at Barnes-Jewish Hospital (BJH), a 1250-bed urban medical center affiliated with Washington University in the St Louis School of Medicine. The study was approved by the Washington University Human Research Protection Office with a waiver of informed consent. VAD recipients were identified through a prospectively collected database that contained the demographic information of all patients who underwent VAD placement at BJH. Patients who expired within 48 hours of VAD implantation or moved to an outside center for follow-up after implantation were not included in the analysis (Figure 1).

Figure 1.

Flow chart of patient attrition. Abbreviations: VAD, ventricular assist device; VADI, ventricular assist device infection.

Data collected from the prospective database included comorbid conditions and VAD implantation information. Data collected by chart review and queries of the BJH Healthcare Medical Informatics database included vital signs, infection-related symptom history, physical examination findings, laboratory and imaging results, as well as antimicrobial and surgical management of infections. The start date of the infection episode was recorded as the first day when symptoms related to the episode were reported by the patient to healthcare personnel, either over the phone or during an in-person visit. For patients in whom workup was initiated by signs or radiographic findings suggestive of infection, rather than patient-reported symptoms, the date when signs were first reported to appear was used as the start date of infection. The presence of systemic inflammatory response syndrome (SIRS) was determined using vital signs and laboratory results first obtained during the encounter when the episode started. SIRS was defined as 2 or more of the following criteria: temperature lower than 96.5°F or higher than 100.4°F, heart rate greater than 90 beats/minute, respiratory rate greater than 20 breaths/minute, and white blood cell (WBC) count less than 4000 or more than 12 000 cells/mm³. Episodes managed as an outpatient with no available vital signs or labs were assigned missing values.

Published ISHLT standardized definitions for infection in patients with mechanical circulatory support include 3 categories: VAD-specific, VAD-related, and non-VAD infections. VAD-specific infections involve the hardware itself or body surfaces containing it, and include infection of the pump or cannula, driveline, or pocket [15]. VAD-related infections are those that can occur in the absence of a device but may be more common in patients with a device, including VAD-related endocarditis, mediastinitis, and bloodstream infections (BSIs). Non-VAD infections are unrelated to the presence of the device (eg, pneumonia, urinary tract infection). Here we analyzed all first episodes of VAD-specific (pump/cannula, pocket, superficial driveline, and deep driveline) and VAD-related (VAD-related endocarditis, mediastinitis, BSI, and BSI-presumed VAD-related) infections as defined by the ISHLT criteria. We excluded central venous catheter–related BSI, non-VAD mediastinitis, and other non-VAD infections from analysis. In patients who fulfilled criteria for more than 1 infection site or ISHLT category, each site/category was included for incidence and time-to-infection calculations. For all other analyses by ISHLT category, patients were considered to have a VAD-specific infection if they had any of the diagnoses in this category; otherwise, they were considered to have a VAD-related infection. Infections were also analyzed as local (driveline or pocket infections, mediastinitis without BSI) or endovascular (pump/cannula infection, infective endocarditis, and BSI), as in previous studies [9]. For these analyses, patients with infection involving multiple VAD sites with the same microorganism were assigned a diagnosis based on the most severe infection (VAD endocarditis > pump/cannula infection > VAD-related or presumed VAD-related BSI > VAD mediastinitis > VAD pocket infection > deep driveline infection > superficial driveline infection). Patients were followed until transplant, death, or end of chart review in January 2017. Antimicrobial suppression was defined as initiation of long-term antimicrobials after completion of the initial course of therapy, with the goal of preventing recurrence. For calculation of suppressive antimicrobials, only patients who survived the completion of the initial antimicrobial course with the VAD in place were considered candidates for suppression.

Statistical Analysis

Descriptive statistics were performed and are reported as counts and percentages for categorical data. For continuous data we calculated means ± standard deviations or medians (ranges), as appropriate. Time-to-event data were analyzed using the Kaplan–Meier technique. P values less than .05 were considered statistically significant. Statistical analyses were performed using SPSS statistics 22.0 (IBM Corporation).

RESULTS

A total of 455 patients who had a continuous-flow VAD implanted from January 2009 to May 2015 were identified. Four patients were excluded from the cohort: 3 due to expiration within 48 hours of VAD implantation and 1 due to transfer of care to another center. The remaining 451 patients were followed for a total 711 person-years of VAD support (range, 0.005–7.5 person-years of VAD support). Median age at the time of implant was 59 years (range, 20–79 years). Baseline patient characteristics and other details about the implanted devices are presented in Tables 2 and 3 (Supplementary Table 4).

Table 2.

Baseline Patient Characteristics; Clinical, Microbiological, and Radiological Features; and Management of Infection for Ventricular Assist Device (VAD)–Specific and VAD-related Infection Categories

	VAD-specific (n = 146)	VAD-related (n = 28)	P Value
Patient characteristics
Male sex, n (%)	117 (80.1)	22 (78.6)	.850
Age at implantation, median (range), years	56 (20–78)	57 (24–72)	.637
BMI at implant, median (range), kg/m2	29.1 (17.7–46.6)	30.2 (20.6–43.9)	.303
CHF type, n (%)
Nonischemic	67 (45.9)	16 (57.1)	.607
Ischemic	72 (49.3)	10 (35.7)
Mixed	4 (2.7)	1 (3.6)
Congenital	3 (2.1)	1 (3.6)
Heartmate II device, n (%)	119 (81.5)	19 (67.9)	.102
Strategy, n (%)
Bridge to transplant	87 (59.6)	18 (64.3)	.490
Destination therapy	52 (35.6)	10 (35.7)
Bridge to decision	7 (4.8)	0
INTERMACS profile <2, n (%)	119 (81.5)	26 (92.9)	.174
Clinical findings,a n/N (%)
Temperature <96.5°F or >100.4°F	21/145 (14.5)	10/27 (37.0)	.011
Heart rate >90 beats/minute	78/145 (53.8)	21/27 (77.8)	.021
Respiratory rate >20 breaths/minute	48/141 (34.0)	20/27 (74.1)	<.001
WBC <4 or >12 000/mm3	46/141 (32.6)	15/27 (55.6)	.023
≥2 SIRS criteria met	56/141 (39.7)	22/27 (81.5)	<.001
Microbiology, n (%)
Gram positives	90 (61.6)	19 (67.9)	.534
Mixed	11 (7.5)	1 (3.6)	.693
Gram negatives	20 (13.7)	7 (25.0)	.154
Fungi	6 (4.1)	0	.591
Culture negative	11 (7.5)	1 (3.6)	.693
No cultures	8 (5.5)	0	.358
Radiographic features on CT,b n (%)
Normal CT	41 (33.1)	12 (66.7)	.001
Driveline-only findings	59 (47.6)	0
Beyond driveline findings	24 (19.4)	6 (33.3)
Medical management
Antimicrobial duration, median (range), days	42 (1–312)	34 (1–173)	.078
Parenteral therapy, n (%)	110 (75.3)	26 (92.9)	.040
Suppressive antimicrobials,c n (%)	98 (73.7)	14 (63.6)	.330
Surgical management, n (%)
Driveline revision or debridement	15 (10.3)	0	.134
Pocket or mediastinal debridement	6 (4.1)	3 (10.7)	.160
Pump or cannula revision	4 (2.7)	0	1.000
LVAD explant	1 (0.7)	1 (3.6)	.297
Other source control, n (%)	8 (5.5)	1 (3.6)	1.000

Abbreviations: BMI, body mass index; CHF, congestive heart failure; CT, computed tomography; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; LVAD, left ventricular assist device; SIRS, systemic inflammatory response syndrome; WBC, white blood cell count.

^aDenominators indicate the number of patients with available data for each parameter.

^bTotal number of CTs’ denominator is 142.

^cIncluded are only patients who survived the completion of the initial antimicrobial course and had VAD in place at the end of said course (n = 155).

Table 3.

Baseline Patient Characteristics; Clinical, Microbiological, and Radiological Features; and Management of Infection for Endovascular and Local Categories

	Endovascular (n = 66)	Local (n = 108)	P Value
Patient characteristics
Male sex, n (%)	54 (81.8)	85 (78.7)	.619
Age at implantation, median (range), years	57 (24–73)	56 (20–78)	.631
BMI at implant, median (range), kg/m2	30.2 (18.4–41.8)	28.7 (17.7–46.6)	.253
CHF type, n (%)
Nonischemic	27 (40.9)	56 (51.9)	.205
Ischemic	33 (50)	49 (45.4)
Mixed	3 (4.5)	2 (1.9)
Congenital	3 (4.5)	1 (0.9)
Heartmate II device, n (%)	51 (77.3)	87 (80.6)	.604
Strategy, n (%)
Bridge to transplant	42 (63.6)	63 (58.3)	.706
Destination therapy	21 (31.8)	41 (38.0)
Bridge to decision	3 (4.5)	4 (3.7)
INTERMACS profile <2, n (%)	58 (87.9)	87 (80.6)	.208
Clinical findings,a n/N (%)
Temperature <96.5°F or >100.4°F	26/65 (40.0)	5/107 (4.7)	<.001
Heart rate >90 beats/minute	44/65 (67.7)	55/107 (51.4)	.04
Respiratory rate >20 breaths/minute	38/65 (58.5)	30/103 (29.1)	<.001
WBC <4 or >12 000/mm3	36/65 (55.4)	25/103 (24.3)	<.001
≥2 SIRS criteria met	46/65 (70.8)	32/103 (31.1)	<.001
Microbiology, n (%)
Gram positives	48 (72.7)	61 (56.5)	.032
Mixed	2 (3.0)	10 (9.3)	.136
Gram negatives	10 (15.2)	17 (15.7)	.917
Fungi	5 (7.6)	1 (0.9)	.030
Culture negative	1 (1.5)	11 (10.2)	.032
No cultures	0	8 (7.4)	.025
Radiographic features on CTb
Normal CT	30 (55.6%)	23 (26.1)	<.001
Driveline-only findings	6 (11.1)	53 (60.2)
Beyond driveline findings	18 (33.3)	12 (13.6)
Medical management
Antimicrobial duration, median (range), days	43 (1–173)	42 (5–312)	.823
Parenteral therapy, n (%)	65 (98.5)	71 (65.7)	<.001
Suppressive antimicrobials,c n (%)	43 (84.3)	69 (66.3)	.019
Surgical management, n (%)
Driveline revision or debridement	2 (3.0)	13 (12.0)	.040
Pocket or mediastinal debridement	3 (4.5)	6 (5.6)	1.000
Pump or cannula revision	4 (6.1)	0	.020
LVAD explant	2 (3.0)	0	.143
Other source control, n (%)	5 (7.6)	4 (3.7)	.302

Abbreviations: BMI, body mass index; CHF, congestive heart failure; CT, computed tomography; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; LVAD, left ventricular assist device; SIRS, systemic inflammatory response syndrome; VAD, ventricular assist device; WBC, white blood cell count.

^aDenominators indicate the number of patients with available data for each parameter.

^bTotal number of CTs’ denominator is 142.

^cIncluded are only patients who survived the completion of the initial antimicrobial course and had VAD in place at the end of said course (n = 155).

A total of 174 (38.6%) patients developed a VAD infection, with an overall incidence of 36.9 cases per 100 person-years of VAD support. Median time to any infection was 150 days (5–2266 days) (Table 1). Cases of infection by year of implant per 100 person-years of support were as follows: 0.57 (2009), 2.67 (2010), 2.81 (2011), 4.64 (2012), 6.43 (2013), 10.96 (2014), 5.66 (2015), and 1.08 (2016). In the 174 infected patients, there were 146 (83.9%) VAD-specific and 28 (16.1%) VAD-related infections. Driveline infection was the most common VAD infection and the most common VAD-specific infection, occurring in 117 patients (67.2%). The most common VAD-related infection was BSI in 61 cases (35%). The median time to VAD-specific infection was 161 days (5–2266) and 81 days (5–2266) for VAD-related infections. Using the endovascular and local definitions, there were 108 (62.1%) local infections and 66 (37.9%) endovascular infections. The most common endovascular infections were VAD-related BSI followed in decreasing order by pump/cannula infection, presumed VAD-related BSI, and endocarditis. The median time to infection in endovascular cases was 88.5 days (5–2266 days) compared with 171.5 days (5–1325 days) in cases with local infections. Incidence and time to infection by infection type are presented in Table 1.

Table 1.

Ventricular Assist Device Infection Incidence and Time to Infection by Infection Type

VAD Infection Category and Diagnosis	Number of Events	Patients, %	Incident Cases per 100 Person-years of VAD Supporta	Time to Infection,a Days
Allb	174	38.6	36.9	150 (5–2266)
VAD-specificb	146	32.4	29.2	161 (5–2266)
Pump/cannula	27	5.9	4.0	71 (20–1305)
Pocket	14	3.1	2.0	66 (19–1101)
Deep driveline	40	8.9	6.1	173.5 (5–2266)
Superficial driveline	78	17.3	13.4	175 (9–981)
VAD-relatedb	72	15.9	11.4	81 (5–2266)
Endocarditis	1	0.2	0.14	83
Bloodstream
VAD-related	47	10.4	7.1	147 (20–2266)
Presumed	14	3.1	2.0	39.5 (5–179)
VAD-related mediastinitis	18	3.9	2.6	36.5 (9–371)

Abbreviation: VAD, ventricular assist device.

^aFor these columns, all distinct infection sites in the 174 patients were included in calculations. Data are presented as medians (ranges).

^bThe number of events in these rows corresponds to the total number of patients with infection within the category as some patients had more than 1 infection site.

SIRS on presentation was more common in patients with VAD-related (81.5%) and endovascular (70.8%) infections as compared with VAD-specific (39.7%) and local (31.1%) infections. Patients with VAD-related and endovascular infections were more likely to have a temperature above 100.4°F or below 96.5°F, WBC count less than 4000 or above 12 000 cells/mm³, and tachycardia than patients with VAD-specific or local infections (Tables 2 and 3). Despite this, 29.2% of patients in the endovascular group did not have SIRS findings on initial evaluation, including the patient with VAD-related endocarditis, 8 cases of pump/cannula infection, and 5 patients with isolated BSIs (Supplementary Figure 3).

Computed tomography (CT) of the chest and abdomen was used for evaluation of the infection site in 142 (81.6%) patients. The most common VAD finding was driveline exit site stranding in 70 patients (49.3%). A drainable fluid collection along the driveline was identified in 15 patients (10.6%). VAD abnormalities beyond the driveline were found in 30 (21.1%) of the patients with CT results. These included stranding, gas, or fluid collections on 1 or more of the following sites: mediastinum (n = 15; 10.6%), pump (n = 6; 4.2%), inflow cannula (n = 7; 4.9%), outflow cannula (n = 14; 9.9%), and device pocket (n = 8; 5.6%).

The etiologic agent(s) of infection were determined in 154 (88.5%) patients (Tables 2 and 3). Gram-positive bacteria were identified in 109 (62.6 %) patients. Of these, methicillin-susceptible Staphylococcus aureus (MSSA; n = 46; 26.5%), methicillin-resistant S. aureus (MRSA; n = 18; 10.3%), and coagulase negative Staphylococci (n = 15; 8.6%) were the most common. MSSA and Enterococci were the predominant gram-positive bacteria causing endovascular infections (n = 14 for both; 21.2% of all endovascular cases). Of the 14 cases of enterococcal endovascular infection, 5 patients (35.7%) had a concurrent episode of gastrointestinal bleed (4 of them also with a central line) and 2 additional patients had a central line as the presumed source of infection. Pseudomonas species were the most common organisms in patients with infections caused by gram-negative bacteria (n = 9; 5.1%), followed by Klebsiella (n = 5; 2.9%) and Enterobacter (n = 4; 2.3%) species. Eight patients (4.6%) with local infections had a clinical diagnosis with no cultures collected. Non–albicans Candida species were isolated in 6 (3.5%) patients, and in all but 1 case were associated with endovascular infection. There were 12 patients diagnosed with infection in which the cultures had no growth, including 1 patient with intraoperative evidence of a pump infection. There were also 12 cases (6.9%) with cultures containing mixed growth. Of these, 9 cultures had 2 or more gram-positive organisms (including MSSA, multiple coagulase-negative Staphylococcus species, and Corynebacterium species), 1 culture had 2 gram-negative organisms (Escherichia coli and Bacteroides fragilis), and 2 cultures had mixed organisms that were not further identified.

Empiric antimicrobial therapy was initiated in 150 patients after initial evaluation. A total of 23 of 24 patients who did not receive empiric treatment had a local infection. There was no standard empiric antimicrobial regimen for suspected VAD infections at our institution during the study period. Oral empiric therapy was used in 23 patients, all with local infection. The most common empiric oral treatments included cephalexin (n = 7), ciprofloxacin (n = 6), linezolid (n = 3), doxycycline (n = 3), and others (amoxicillin, clindamycin, trimethoprim/sulfamethoxazole). The remaining 127 patients received parenteral empiric antibiotic treatment. Parenteral empiric regimens were varied but were predominantly composed of an agent to cover resistant gram-positive organisms, alone or in combination with an antipseudomonal agent. Most frequent regimens used were vancomycin plus cefepime (n = 73) and vancomycin alone (n = 19). Empiric antifungal coverage with micafungin was provided for 7 patients. Of the 150 patients who received an empiric antibiotic regimen, 132 (88%) had culture data available to assess the appropriateness of the empiric regimen. Coverage for the cultured organism(s) was included in the empiric regimen in 121 patients (91.7%). The empiric regimen was narrowed or changed upon receipt of culture results in 116 patients (87.9%). Antimicrobial therapy was used to treat all 174 patients with infection. The median duration of therapy was 42 days for VAD-specific versus 34 days for VAD-related infections (P = .078) (Table 2). For endovascular and local infections, the median duration was approximately 6 weeks (median: 43 vs 42 days respectively; P = .823) (Table 3). Patients with VAD-related and endovascular infections were more likely to receive a parenteral antimicrobial regimen for definitive treatment (92.9% and 98.5%, respectively) than patients with VAD-specific or local (75.3% and 65.7%, respectively) infections (Tables 2 and 3). Of the 155 patients who survived completion of their initial antimicrobial regimen, 112 (72.2%) were started on antimicrobial suppression. Suppressive antimicrobials were more common in the endovascular infection group than in patients with local infections (84.3% vs 66.3%; P = .019).

For management of infection, 39 (22.4%) patients underwent surgical intervention in addition to antimicrobial therapy. The most common surgical procedures were driveline revision in 15 and mediastinal and/or pocket debridement in 9 patients (Tables 2 and 3).

Median survival from the date of the initial infection episode was 28 months. Median survival from infection was significantly lower in patients with VAD-related compared with VAD-specific infections (14 vs 35 months) (P = .007) (Figure 2A). The median survival from the time of infection was significantly lower for patients with endovascular infection (median, 14 months) compared with patients with local infection (median, 35 months) (P < .001) (Figure 2B).

Figure 2.

A, Survival from the time of infection in patients with VAD-specific and VAD-related infection. B, Survival from the time of infection in patients with endovascular and local infection. Abbreviations: Endo, endovascular; VAD, ventricular assist device.

DISCUSSION

This study includes the largest nonregistry cohort of patients followed up to a first episode of infection related to a VAD using standardized definitions for infection. Overall, the incidence and epidemiology of infection remain similar to reports from the early era of continuous-flow devices, despite increased experience with surgical implantation and additional strategies for device stabilization and care. Studies that included patients with continuous-flow VADs implanted from 2005–2011 and 2006–2009 reported rates of infection of 32 to 35.5 per 100 person-years, respectively [9, 18], comparable to our incidence of 36.9 infections per 100 person-years. Infection incidence initially increased by year at our institution, with a later decrease in incidence, perhaps due to increased experience with techniques for implantation and protocol development for VAD care.

The most frequent infection continued to be of the device driveline, diagnosed in 25.9% of the VAD cohort. Studies in the early era of continuous-flow devices had significant variability in their reported rates of driveline infection, ranging from 17% to 48% of all patients with a VAD [19, 20], likely due to variability in the definitions of infection, studying selected populations (limited to transplant patients, late-onset infections), etc. More recent reports using standardized definitions of infection have reported incidence rates of driveline infection ranging from 16.1% to 23.8% [9, 21], consistent with our finding of a 25.9% incidence.

In 2017, the ISHLT published a consensus document with recommendations targeted towards the evaluation and clinical management of VAD infections [17]. One of these recommendations is to consider outpatient management of patients presenting with symptoms limited to the driveline and no signs of sepsis. Although fever and leukocytosis appear to be discriminants of endovascular versus local infections, our data caution against a rigorous application of the recommendation, as approximately 30% of the patients with endovascular infection did not meet SIRS criteria at the onset of infection.

The ISHLT recommendations also state that currently no imaging modality can definitively exclude deep-tissue-space infections but recommend considering CT or ultrasound imaging for this purpose. The number of normal CTs in our cohort may reflect the lack of sensitivity of CT imaging, as 37.7% of the infected patients who underwent CT had normal imaging. Patients with findings beyond the driveline in our cohort were more likely to present without driveline-related symptoms and were more likely to have tachycardia and leukocytosis on initial workup.

The microbiology of infection continues to be predominantly due to gram-positive bacteria, particularly Staphylococci. MRSA was identified in 10.8% of the infected patients with a culture performed, whereas MSSA was identified in 27.7% of the cultured infected patients. It is difficult to draw conclusions regarding any trends in MRSA and multidrug-resistant infections in VAD patients due to a lack of detail in previously reported microbiologic data [9, 18] and the possibility of regional differences in antimicrobial resistance [8].

Interestingly, the 14 enterococcal infections observed in our cohort were limited to endovascular sites. We considered central lines, gastrointestinal bleeding, and bacterial translocation as potential risk factors for this phenomenon; however, half of the patients with enterococcal endovascular infection did not have a concurrently diagnosed gastrointestinal bleed or a central line in place.

The median number of antimicrobial days in our cohort was over 4 weeks, independent of infection category or type. Although ISHLT recommendations suggest there may be a role for long-term suppressive antimicrobials, there are limited data to support this intervention. Antimicrobial suppression was prescribed in 112 (72.2%) of the eligible patients in our cohort. Additional studies evaluating benefits and risks of antimicrobial suppression strategies for the long-term management of patients with VAD infections are greatly needed, especially with the expected increase in the number of patients on destination therapy and potential for infectious complications in the long term.

Given its retrospective nature our study has inherent associated limitations, including the potential for observational bias and misclassification. To minimize these biases, we used predefined and standardized definitions for the diagnosis of infection and for other data collected. Because of local practices many of our patients with BSIs did not undergo trans-esophageal echocardiography, which could cause misclassification of endocarditis cases; however, we did not observe an increased rate of pump/cannula infections, VAD-related, and presumed VAD-related BSIs compared with prior reports [9], a finding that would be expected if endocarditis cases had been shifted to an alternative diagnosis.

In conclusion, VAD infections are a significant problem in the day-to-day management of patients with VADs despite the advances in the field. Studies linking clinical parameters and diagnostic testing may provide information to plan practical workup strategies in these patients. VAD infections are very heterogeneous, and future studies should target specific entities to better characterize their presentation and guide their management. Studies looking at the role of long-term antimicrobials and infection relapse in VAD patients are greatly needed so that decisions regarding the benefits and risks of these interventions are better informed.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The authors acknowledge contributions from Kimberly Reske and Cherie Hill for their assistance with data retrieval through medical informatics at Barnes-Jewish Hospital.

Disclaimer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or Washington University School of Medicine.

Financial support. This work was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health (grant number UL1 TR002345). Redcaps was used for the development and maintenance of the study database with the support of a Clinical and Translational Science Award (grant number UL1 TR000448) and the Siteman Comprehensive Cancer Center Award (grant number P30 CA091842) from Washington University School of Medicine.

Potential conflicts of interest. M. A. O. reports personal fees from Pfizer and grants from Pfizer, Merck, and Sanofi Pasteur outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.