2018 Volume 82 Issue 2 Pages 448-456
Background: Blood stream infection is thought to increase the risk of hemorrhagic stroke, a major adverse event with devastating outcome, in patients with continuous-flow left ventricular assist devices (LVADs). We analyzed the risk factors of hemorrhagic stroke in LVAD patients, as well as the time relationship between systemic bacteremia and hemorrhagic stroke.
Methods and Results: We evaluated the incidence of systemic bacteremia and stroke in 164 patients who underwent continuous-flow LVAD implantation between 2005 and 2016. At 1 and 2 years after implantation, the incidence of bacteremia was 29% and 36%, and the incidence of hemorrhagic stroke was 22% and 22% in patients without bacteremia, and 32% and 44% in those with bacteremia, respectively (P=0.035). This higher prevalence of hemorrhagic stroke in patients with bacteremia was notable particularly in the chronic phase (>90 days after implantation). Multivariate analysis revealed that bacteremia was an independent risk factor of hemorrhagic stroke in the chronic phase [hazard ratio, 2.36 (1.02–5.62); P=0.044]. The hazard rate was the highest immediately after the onset of bacteremia, and the risk steadily declined by 90 days after the last episode of bacteremia and flattened thereafter.
Conclusions: Bacteremia was an independent risk factor of hemorrhagic stroke in patients in the chronic phase, with the highest risk seen in the early phase following an episode of bacteremia.
Implantable continuous-flow left ventricular assist devices (LVADs) have become a prevalent treatment option for endstage heart failure patients.1 Although a continuous-flow LVAD has a lower rate of stroke as compared with older pulsatile LVADs, stroke remains a major adverse event.2,3 Hemorrhagic stroke can deteriorate neurological function to a greater degree than ischemic stroke,3 but studies of risk factors for hemorrhagic stroke are limited.4 Reports have noted that bloodstream infection increased the risk of hemorrhagic stroke, though timing was not evaluated,5,6 and the relationship between the time of blood stream infection and subsequent hemorrhagic stroke remains unknown. In the present study we analyzed the risk factors of hemorrhagic stroke following LVAD implantation, and the time relationship between systemic bacteremia and hemorrhagic stroke.
The institutional ethical committee approved this study. Between January 2005 and December 2016, 164 Japanese patients underwent continuous-flow LVAD implantation for endstage heart failure. All patients and family members provided informed consent to participate in related clinical studies. Follow-up data were collected for 0–4.84 years [mean, 1.70±1.12 years; median, 1.54 years; interquartile range (IQR), 0.70–2.62 years)]. All follow-up examinations were completed by April 30, 2017.
Devices and Surgical ProceduresThe type of LVAD utilized was decided at the preoperative conference, though available devices have changed over time in Japan. Briefly, the DuraHeart (Terumo Heart, Ann Arbor, MI, USA) and EVAHEART (SunMedical, Japan) were approved in October 2011, the HeartMateII (Thoratec, Pleasanton, CA, USA) in April 2013, and the Jarvik2000 (Jarvik Heart, NY, USA) in May 2014, while the HeartWare (HeartWare, Framingham, MA, USA) is presently only available for clinical trial use.
Generally, patients underwent LVAD implantation through a median sternotomy with systemic cardiopulmonary bypass, during which the inflow cannula was inserted into the left ventricle apex and the outflow graft was anastomosed to the ascending aorta; 3 patients underwent Jarvik2000 implantation through a left thoracotomy with a descending aorta anastomosis of the outflow graft because of previous cardiac surgery. In patients with mild or greater aortic insufficiency, aortic valve closure with a bovine pericardial patch or replacement with a tissue valve was performed, and mitral valve repair was done for severe functional mitral regurgitation at the surgeon’s discretion. We typically perform a tricuspid valve procedure for moderate or greater tricuspid regurgitation. The decision regarding concomitant right ventricular assist device (RVAD) insertion was made by the attending surgeon.
Anticoagulant and Medical Therapy During LVAD SupportFollowing implantation, all patients received a standardized heart failure prevention medical regimen. Antiplatelet therapy with aspirin (100 mg/day) was also given, starting from the day after implantation and maintained throughout LVAD support, unless major bleeding occurred. In this study period, no patients took dual antiplatelet therapy for prevention of thrombosis. Anticoagulation therapy with warfarin was also initiated, because adequate hemostasis is complemented and maintained at the target international normalized ratio (INR) of 2.0–2.5 for patients with a HeartMateII or DuraHeart, and 2.5–3.0 for those with a HeartWare, EVAHEART, or Jarvik2000. Systemic heparinization was managed until target INR was achieved. Perioperative antibiotic therapy was routinely managed by experienced intensive care unit physicians.
Definition and Management of LVAD-Related Adverse EventsA stroke event was defined as a neurological symptom lasting >24 h with a compatible new lesion shown by brain computed tomography (CT), then subdivided into ischemic and hemorrhagic stroke events. Hemorrhagic stroke included cerebral, cerebellar or subarachnoid hemorrhage. Ischemic stroke was defined as a cerebral or cerebellar infarction detected by brain CT without hemorrhagic lesions. Cerebral infarction accompanied by hemorrhagic transformation lesion was included as an ischemic stroke. Upon occurrence of LVAD-related stroke, laboratory testing was always immediately performed to examine the patient’s level of anticoagulation, the patient was hospitalized and treatment was based on discussion among experienced cardiologists, cardiovascular surgeons, and neurologists. Cessation of antiplatelet and/or anticoagulation therapy was also managed after discussion. Basically, when the patients had any type of intracranial hemorrhagic lesion, antiplatelet and anticoagulant therapy were immediately ceased and factor IX or fresh frozen plasma was started for complete normalization of the INR, and closed consecutive follow-up brain CT was performed until hemostasis was confirmed. Indications for selective angiography or intracranial aneurysm treatment were discussed with the neurosurgeons. After the confirmation of complete hemostasis by serial brain CT, antiplatelet and/or anticoagulation therapy was carefully re-initiated. The therapeutic-target INR after stroke was adjusted by each attending physician according to the patient’s condition.
When an LVAD patient develops a fever >38.0℃, we usually examine blood cultures, and investigate the source of infection using CT or gallium-scintigraphy. An episode of systemic bacteremia was defined as any positive blood culture. For patients who were highly suspected to have active LVAD-related infection, blood cultures were examined regardless of fever. In addition, for patients being treated with antibiotics for definitive LVAD-related infection, blood cultures were examined regularly to confirm the absence of systemic infection and sensitivity of antibiotics. Patients with recurrence of bacteremia were defined as patients with multiple episodes of positive blood cultures during LVAD support. Multiple positive blood culture tests on consecutive days during 1 episode of infection were not defined as recurrence of bacteremia. Management of the driveline exit site was done primarily by experienced surgeons and nurses. For patients with any type of infection, the antibiotic protocol was usually decided by an infection control team physician. When surgical treatment for infection was required, aggressive debridement using a vacuum-assisted closure device without delay is the preferred protocol at our institution.
Data CollectionAll patient data were obtained from electronic medical records and included baseline characteristics, comorbidities, and LVAD type and implantation duration. Major adverse events requiring readmission during LVAD support were also recorded, including cerebral events and LVAD-related infection. All patients were routinely followed, so nearly all of the relevant data were available.
Statistical AnalysisAll statistical analyses were performed using SPSS version 23.0 (IBM®, NY, USA) Categorical variables are summarized as frequencies and percentages, and were compared among groups using a chi-square or Fisher’s exact test. Continuous variables are summarized as the mean±standard deviation or median (IQR). All P-values for statistical analysis were 2-tailed and P<0.05 indicated a statistically significant difference.
In the analysis of the association between hemorrhagic stroke and bacteremia, 7 patients who developed an initial bacteremia episode after any stroke episode were excluded to avoid original contained bias. The incidence rates of stroke and bacteremia were estimated using Kaplan-Meier curves, and compared among the groups using log-rank tests. Cox regression hazard models were used to identify risk factors for bacteremia and hemorrhagic stroke. Baseline characteristics, laboratory data, hemodynamics, intraoperative parameters, and postoperative complications were used as factors. Initially, univariate analysis was applied, then factors with a P-value <0.05 were considered for a multivariate Cox hazard model to identify risk factors. To investigate risk, the odds ratio (OR) of bacteremia-related hemorrhagic stroke, after multivariable adjustments by all covariates’ logistic regression using the marginal structural models of propensity score method, was applied. As covariates, sex, age (decade), and type of LVAD, logarithm of serum C-reactive protein and lactate dehydrogenase level at 3 months after LVAD implantation were selected to determine the inverse propensity score. Time-dependent hazard rate of hemorrhagic stroke at the duration after the last episode of bacteremia was analyzed by Epanechnikov Kernel-smoothed density estimation.
Preoperative characteristics of the 164 enrolled patients are shown in Table 1. Mean age was 42±14 years and 111 (68%) were men. Because of the strict requirements for heart transplantation list eligibility, the comorbidity prevalence was low. The majority of patients had idiopathic-dilated cardiomyopathy and 98% of the entire cohort underwent LVAD implantation as bridge-to-transplantation therapy, because destination therapy had not been approved in Japan at the end of March 2017.
Overall (n=164) |
ICH (−) (n=119) |
ICH (+) (n=45) |
P value | Bacteremia (−) (n=106) |
Bacteremia (+) (n=58) |
P value | |
---|---|---|---|---|---|---|---|
Age (years) | 42±14 | 42±14 | 43±17 | 0.646 | 43±15 | 42±14 | 0.686 |
Male, n (%) | 111 (68) | 84 (71) | 27 (60) | 0.261 | 68 (64) | 43 (74) | 0.303 |
Etiology | |||||||
Ischemic CM, n (%) | 20 (12) | 14 (12) | 6 (13) | 0.792 | 12 (11) | 8 (14) | 0.804 |
Idiopathic DCM, n (%) | 95 (58) | 71 (60) | 24 (53) | 0.483 | 64 (60) | 31 (53) | 0.325 |
Hypertrophic CM, n (%) | 23 (14) | 15 (13) | 8 (18) | 0.451 | 13 (12) | 10 (17) | 0.484 |
Other, n (%) | 25 (16) | 18 (15) | 7 (16) | 0.953 | 15 (14) | 10 (17) | 0.649 |
Comorbidities | |||||||
Diabetes mellitus, n (%) | 24 (15) | 18 (15) | 6 (13) | 1.000 | 18 (17) | 6 (10) | 0.258 |
Hypertension, n (%) | 12 (7) | 8 (7) | 4 (9) | 0.738 | 5 (5) | 7 (12) | 0.120 |
Previous smoking, n (%) | 59 (36) | 44 (37) | 15 (33) | 0.718 | 33 (31) | 26 (45) | 0.128 |
Episode of stroke, n (%) | 10 (6) | 9 (8) | 1 (2) | 0.288 | 5 (5) | 5 (9) | 0.498 |
Preoperative hemodynamics | |||||||
Inotropes, n (%) | 145 (88) | 105 (88) | 40 (89) | 1.000 | 96 (91) | 49 (84) | 0.130 |
IABP, n (%) | 30 (18) | 24 (20) | 6 (13) | 0.371 | 18 (17) | 12 (21) | 0.675 |
Intubation, n (%) | 29 (18) | 20 (17) | 9 (20) | 0.650 | 17 (16) | 12 (21) | 0.527 |
Extracorporeal LVAD, n (%) | 29 (18) | 20 (17) | 9 (20) | 0.650 | 12 (11) | 17 (29) | 0.010 |
INTERMACS profile | 0.515 | 0.018 | |||||
I, n (%) | 12 (7) | 7 (6) | 5 (11) | 7 (7) | 5 (9) | ||
II, n (%) | 59 (36) | 48 (40) | 11 (24) | 39 (37) | 20 (34) | ||
III, n (%) | 59 (36) | 41 (34) | 18 (40) | 44 (42) | 15 (26) | ||
IV, n (%) | 5 (3) | 3 (3) | 2 (4) | 3 (3) | 2 (3) | ||
Bridge to bridge, n (%) | 29 (18) | 20 (17) | 9 (20) | 12 (11) | 17 (29) | ||
LVAD | 0.018 | 0.009 | |||||
HeartMateII, n (%) | 53 (32) | 43 (36) | 10 (22) | 41 (39) | 12 (21) | ||
DuraHeart, n (%) | 38 (23) | 32 (27) | 6 (13) | 23 (22) | 15 (26) | ||
Jarvik2000, n (%) | 33 (20) | 18 (15) | 15 (33) | 24 (23) | 9 (16) | ||
EVAHEART, n (%) | 25 (15) | 15 (13) | 10 (22) | 10 (9) | 15 (26) | ||
HeartWare, n (%) | 15 (9) | 11 (9) | 4 (9) | 7 (7) | 8 (14) | ||
BTT, n (%) | 160 (98) | 117 (98) | 43 (96) | 0.303 | 103 (97) | 57 (98) | 0.619 |
Destination therapy, n (%) | 4 (2) | 2 (2) | 2 (4) | 2 (2) | 2 (3) | ||
CVP (mmHg) | 8.1±4.7 | 7.8±4.9 | 8.8±5.1 | 0.362 | 7.7±4.6 | 9.0±5.6 | 0.149 |
mPAP (mmHg) | 28.2±10.3 | 27.7±9.5 | 29.8±11.4 | 0.308 | 28.4±9.9 | 27.8±10.5 | 0.773 |
PCWP (mmHg) | 20.4±8.3 | 19.7±7.9 | 22.3±9.3 | 0.148 | 20.2±7.7 | 20.7±9.7 | 0.751 |
Cardiac index (L/min/m2) | 2.14±0.59 | 2.18±0.62 | 2.04±0.49 | 0.271 | 2.14±0.58 | 2.15±0.62 | 0.877 |
LVDd (mm) | 68.6±14.4 | 70.0±12.9 | 65.4±13.7 | 0.065 | 69.4±13.8 | 66.3±14.8 | 0.197 |
LVDs (mm) | 63.1±14.9 | 63.9±14.5 | 59.9±14.9 | 0.140 | 64.8±12.6 | 59.2±17.6 | 0.029 |
LVEF (%) | 21.1±8.8 | 20.7±8.2 | 20.9±10.4 | 0.911 | 19.7±8.2 | 22.8±9.6 | 0.048 |
Laboratory data | |||||||
AST (IU/dL) | 27 (21–36) | 27 (21–36) | 26 (22–34) | 0.480 | 26 (20–36) | 29 (24–36) | 0.641 |
ALT (IU/dL) | 23 (15–37) | 24 (17–38) | 21 (14–34) | 0.207 | 23 (15–37) | 23 (16–38) | 0.622 |
Total bilirubin (mg/dL) | 1.0 (0.6–1.5) | 1.0 (0.7–1.5) | 0.8 (0.6–1.5) | 0.248 | 1.1 (0.7–1.7) | 1.0 (0.6–1.4) | 0.375 |
Albumin (mg/dL) | 3.7±0.7 | 3.7±0.7 | 3.6±0.7 | 0.524 | 3.7±0.6 | 3.5±0.7 | 0.027 |
Blood urea nitrogen (mg/dL) | 21±12 | 19±10 | 24±15 | 0.029 | 20±11 | 22±13 | 0.265 |
Creatinine (mg/dL) | 0.99 (0.73–1.26) |
0.97 (0.73–1.24) |
1.02 (0.75–1.39) |
0.666 | 0.99 (0.72–1.26) |
1.02 (0.73–1.26) |
0.745 |
Ccr (mL/min/1.73 m2) | 66±32 | 68±32 | 62±32 | 0.328 | 66±31 | 68±35 | 0.652 |
PT-INR | 1.63±0.64 | 1.63±0.64 | 1.62±0.67 | 0.933 | 1.56±0.59 | 1.75±0.72 | 0.092 |
WBC count (×103/μL) | 6.8±2.8 | 6.7±2.5 | 7.0±3.5 | 0.532 | 6.8±3.0 | 6.8±2.6 | 0.991 |
Hemoglobin (g/dL) | 11.5±1.9 | 11.5±1.9 | 11.4±1.7 | 0.749 | 11.5±1.8 | 11.4±2.0 | 0.691 |
ALT, alanine transaminase; AST, aspartate transaminase; BTT, bridge-to-transplantation; Ccr, creatinine clearance; CM, cardiomyopathy; CVP, central venous pressure; DCM, dilated cardiomyopathy; IABP, intra-aortic balloon pumping; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; LVAD, left ventricular assist device; LVDd, left ventricular end-diastolic dimension; LVDs, left ventricular end-systolic dimension; LVEF, left ventricular ejection fraction; mPAP, mean pulmonary artery pressure; PCWP, pulmonary capillary wedge pressure; PT-INR, prothrombin time international normalization rate; WBC, white blood cell.
In total, 45 patients had 60 hemorrhagic stroke events (0.21 events/patient-year) and 15 patients developed 20 ischemic stroke events (0.07 events/patient-years) during all follow-up periods. The incidence of any stroke and stroke-related death in the entire cohort is shown in Figure 1A. The incidence rate of stroke at 1 and 2 years was 33% and 41%, and of stroke-related death was 11% and 13%, respectively. After dividing stroke cases into hemorrhagic and ischemic stroke, the incidence rate of hemorrhagic stroke at 1 and 2 years was 26% and 31%, and of ischemic stroke was 10% and 15%, respectively (Figure 1B). During the follow-up period, 45 (27%) patients had 65 hemorrhagic stroke events. The mean prothrombin time international normalization rate (PT-INR) at the time of their first stroke event was 2.49±0.67 in hemorrhagic stroke, and 2.40±0.69 in ischemic stroke (P=0.676). The mean platelet count at the time of first stroke events was 193±93×103/μL in hemorrhagic stroke, and 169±67×103/μL in ischemic stroke (P=0.411). There were 9 patients who underwent decompressive craniectomy after hemorrhagic stroke, and 1 patient underwent selective angiography for suspected mycotic aneurysm rupture. However, only 2 (20%) of these 10 patients with an intracranial procedure survived after the hemorrhagic stroke. In total, 16 patients (10%) died from hemorrhagic stroke after a median support period of 144 (54–334) days, while 2 died of ischemic stroke after 388 and 1,257 days, respectively. The incidence of hemorrhagic stroke-related death at 1 and 2 years was 9% and 11%, and that of ischemic stroke death was 0% and 1%, respectively.
(A) Rates of incidence of overall stroke (solid line) and stroke-related death (dashed line) in the entire cohort. (B) Stroke was divided into hemorrhagic (solid red line) and ischemic (solid green line). Rates of hemorrhagic stroke-related (dashed red line) and ischemic stroke-related (dashed green line) death are also shown. LVAD, left ventricular assist device.
The rates of bacteremia incidence are shown in Figure 2. At 1 and 2 years, incidence was 29%, and 36%, respectively. During the follow-up period, 58 patients (35%) had at least 1 episode of bacteremia, with a median duration from LVAD implantation to first episode of 124 (29–433) days. Staphylococcus aureus was detected in 18 patients, including 9 with methicillin-resistant S. aureus; Pseudomonas aeruginosa was detected in 9 patients. Of these 58 patients, 30 (52%) had recurrent bacteremia. A driveline exit site infection occurred in 38 (23%) patients, of whom 11 required surgical debridement of the driveline, and 4 underwent LVAD exchange because of severe driveline infection. LVAD pump pocket infection occurred in 12 (7%) patients, 6 of whom were bridged from an extracorporeal LVAD, while another 2 cases were caused by an invasive driveline infection. Cox hazard analysis revealed a requirement of postoperative continuous venovenous hemodialysis as a risk factor for bacteremia development (Table 2).
Rates of incidence and 95% confidence interval of bacteremia in all 164 study patients. LVAD, left ventricular assist device.
Univariate | Multivariate | |||
---|---|---|---|---|
OR (95% CI) | P value | OR (95% CI) | P value | |
Age (years) | 1.00 (0.98–1.02) | 0.853 | ||
Male | 1.28 (0.73–2.34) | 0.399 | ||
BMI (kg/m2) | 1.02 (0.93–1.11) | 0.691 | ||
BSA (0.1 m2) | 1.00 (0.91–1.23) | 0.984 | ||
Ischemic etiology | 1.36 (0.60–2.71) | 0.437 | ||
DCM | 0.69 (0.41–1.15) | 0.148 | ||
LVDd (mm) | 0.98 (0.96–1.00) | 0.016 | 1.00 (0.97–1.02) | 0.817 |
LVEF (%) | 1.03 (1.00–1.06) | 0.082 | ||
ECMO | 1.93 (0.58–4.76) | 0.251 | ||
IABP | 1.81 (0.94–3.26) | 0.076 | ||
CVP (mmHg) | 1.05 (0.98–1.12) | 0.137 | ||
mPAP (mmHg) | 0.99 (0.96–1.02) | 0.534 | ||
PCWP (mmHg) | 1.00 (0.96–1.04) | 0.990 | ||
Cardiac index (L/min/m2) | 1.09 (0.62–1.85) | 0.759 | ||
Hemoglobin (g/dL) | 0.96 (0.82–1.01) | 0.360 | ||
WBC count (×103/μL) | 1.01 (0.92–1.10) | 0.803 | ||
C-reactive protein (mg/dL) | 1.08 (1.01–1.14) | 0.032 | 1.00 (0.86–1.15) | 0.964 |
Serum creatinine (mg/dL) | 0.99 (0.78–1.04) | 0.806 | ||
Albumin (mg/dL) | 0.51 (0.34–0.76) | 0.001 | 0.64 (0.31–1.34) | 0.231 |
Serum total bilirubin (mg/dL) | 0.98 (0.82–1.02) | 0.507 | ||
Redo surgery | 1.00 (0.56–1.72) | 0.996 | ||
Conversion from extracorporeal LVAD | 2.34 (1.29–4.05) | 0.006 | 0.88 (0.28–2.39) | 0.820 |
LVADs without pump pocket | 0.98 (0.54–1.70) | 0.951 | ||
Use of EVAHEART* | 2.36 (1.26–4.16) | 0.005 | 1.86(0.80–4.09) | 0.242 |
Operative time (h) | 1.00 (0.84–1.18) | 0.970 | ||
CPB time (h) | 1.05 (0.77–1.42) | 0.754 | ||
RCC transfusion (units) | 1.03 (1.00–1.05) | 0.074 | ||
RVAD requirement | 2.56 (1.28–4.76) | 0.009 | 1.69 (0.52–4.63) | 0.335 |
Re-exploration for bleeding | 1.15 (0.53–2.28) | 0.709 | ||
Postoperative CVVHD | 4.84 (2.28–9.85) | <0.001 | 2.82 (1.03–7.49) | 0.045 |
Ventilator support duration (days) | 1.01 (0.93–1.07) | 0.723 |
*Other LVADs were not a risk factor in the univariate analysis. BMI, body mass index; BSA, body surface area; CI, confidence interval; CPB, cardiopulmonary bypass; CVVHD, continuous venovenous hemodialysis; ECMO, extracorporeal membrane oxygenation; OR, odds ratio; RCC, red cell concentrate; RVAD, right ventricular assist device. Other abbreviations as in Table 1.
For the relationship between bacteremia and hemorrhagic stroke, we analyzed the incidence of hemorrhagic stroke according to bacteremia episode during LVAD support (Figure 3A). The overall incidence of hemorrhagic stroke at 1 and 2 years was, respectively, 22% and 22% in patients without any episode of bacteremia, but 32% and 44% in patients with an episode of bacteremia (P=0.035). Because the majority of hemorrhagic stroke incidents in patients without any bacteremic episode occurred within 90 days, we performed a subset analysis of hemorrhagic stroke that occurred during the chronic phase (>90 days) (Figure 3B). That showed an overall incidence of hemorrhagic stroke after >90 days of LVAD support in patients without bacteremia at 1 and 2 years of 7% and 7%, respectively, and 19% and 34% in patients with bacteremia (P<0.001). In patients without a bacteremic episode, no hemorrhagic strokes occurred after 1 year of support.
(A) Incidence of hemorrhagic stroke according to an episode of bacteremia during LVAD support (7 patients with an initial episode of bacteremia after a stroke event were excluded from analysis to avoid contained bias). (B) Rates of incidence of hemorrhagic stroke >90 days after LVAD implantation (blue: patients without bacteremia, red: patients with bacteremia). LVAD, left ventricular assist device.
For determining the risk factors of hemorrhagic stroke in the acute (≤90 days after LVAD implantation) and chronic (>90 days) phases, we performed Cox hazard regression analysis. Multivariate analysis revealed older age as an independent risk factor for hemorrhagic stroke in the acute phase. A bacteremic episode during the perioperative period was not a risk factor for hemorrhagic stroke in the acute phase (Table 3) The analysis for risk factors of hemorrhagic stroke in the chronic phase is shown in Table 4. Bacteremia during the acute phase, as well as higher C-reactive protein and serum lactate dehydrogenase levels at 3 months, was an independent risk factor for the incidence of hemorrhagic stroke in the chronic phase. We also performed multivariate analysis with adjustment by covariates’ logistic regression using the marginal structural models of the propensity score method to evaluate the effect of each LVAD type on hemorrhagic stroke. As covariates, sex, age, type of LVAD, and episode of bacteremia, logarithm of serum C-reactive protein and serum lactate dehydrogenase at 3 months after LVAD implantation were selected to weigh the inverse propensity score. Type of LVAD was not a risk factor, but an episode of bacteremia was the only significant risk factor for hemorrhagic stroke (OR: 2.22, range 1.40–3.53; P<0.001). When we further performed stratified model analysis for each type of LVAD, no type of device contributed to the risk of hemorrhagic stroke (Table S1).
Univariate | Multivariate | |||
---|---|---|---|---|
HR (95% CI) | P value | HR (95% CI) | P value | |
Age (years) | 1.05 (1.01–1.09) | 0.010 | 1.08 (1.02–1.15) | 0.003 |
Male | 0.49 (0.20–1.23) | 0.126 | ||
BMI (kg/m2) | 0.94 (0.80–1.09) | 0.444 | ||
BSA (0.1 m2) | 0.96 (0.85–1.14) | 0.566 | ||
Ischemic etiology | 1.74 (0.50–4.66) | 0.347 | ||
DCM | 0.88 (0.48–2.65) | 0.759 | ||
Diabetes mellitus | 0.94 (0.28–2.41) | 0.903 | ||
Hypertension | 1.50 (0.36–4.26) | 0.528 | ||
Episode of stroke | 0.52 (0.03–2.44) | 0.475 | ||
LVDd (mm) | 0.98 (0.96–1.01) | 0.183 | ||
LVEF (%) | 0.98 (0.93–1.04) | 0.541 | ||
CVP (mmHg) | 1.03 (0.93–1.13) | 0.593 | ||
mPAP (mmHg) | 1.00 (0.95–1.04) | 0.935 | ||
PCWP (mmHg) | 1.01 (0.95–1.07) | 0.827 | ||
IABP | 0.97 (0.23–2.86) | 0.967 | ||
ECMO | 2.65 (0.36–7.76) | 0.323 | ||
Hemoglobin (g/dL) | 0.80 (0.61–1.01) | 0.079 | ||
WCC count (×1,000/μL) | 1.12 (0.96–1.26) | 0.145 | ||
C-reactive protein (mg/dL) | 1.06 (0.95–1.16) | 0.264 | ||
Serum creatinine (mg/dL) | 1.13 (0.75–1.38) | 0.451 | ||
Albumin (mg/dL) | 0.64 (0.34–1.21) | 0.164 | ||
Serum total bilirubin (mg/dL) | 0.98 (0.64–1.02) | 0.618 | ||
Redo surgery | 1.98 (0.79–4.91) | 0.143 | ||
Conversion from extracorporeal LVAD | 1.06 (0.30–2.83) | 0.923 | ||
Operative time (h) | 1.18 (0.93–1.46) | 0.169 | ||
CPB time (h) | 1.98 (1.23–3.22) | 0.005 | 1.43 (0.84–2.39) | 0.184 |
RCC transfusion (units) | 1.02 (0.96–1.06) | 0.462 | ||
Aortic valve surgery | 1.73 (0.27–6.02) | 0.497 | ||
Axial-flow device | 1.76 (0.76–4.37) | 0.187 | ||
Re-exploration for bleeding | 1.43 (0.51–3.46) | 0.473 | ||
RVAD requirement | 0.34 (0.02–1.62) | 0.210 | ||
Inotropic support duration (days) | 1.00 (0.98–1.01) | 0.972 | ||
CVVHD requirement | 3.03 (1.06–7.83) | 0.040 | 2.17 (0.56–7.07) | 0.240 |
Bacteremia during acute phase | 1.29 (0.43–3.23) | 0.624 | ||
Post-LVAD LVDd (mm) | 0.98 (0.95–1.01) | 0.125 |
CVVHD, continuous venovenous hemodialysis; HR, hazard ratio. Other abbreviations as in Tables 1,2.
Univariate | Multivariate | |||
---|---|---|---|---|
HR (95% CI) | P value | HR (95% CI) | P value | |
Age (years) | 0.99 (0.97–1.02) | 0.698 | ||
Male | 0.72 (0.34–1.63) | 0.420 | ||
BMI (kg/m2)* | 0.83 (0.72–0.96) | 0.057 | ||
BSA (0.1 m2)* | 0.79 (0.63–1.01) | 0.056 | ||
Ischemic etiology | 0.73 (0.12–2.44) | 0.647 | ||
DCM | 0.60 (0.28–1.28) | 0.184 | ||
Diabetes mellitus | 0.82 (0.13–2.95) | 0.793 | ||
Hypertension | 1.09 (0.06–5.38) | 0.936 | ||
Axial-flow device | 1.20 (0.56–2.54) | 0.635 | ||
Device without pump pocket | 2.20 (1.02–4.63) | 0.046 | ||
Redo surgery | 0.60 (0.17–1.64) | 0.339 | ||
Conversion from extracorporeal LVAD | 1.20 (0.34–3.25) | 0.749 | ||
Re-exploration for bleeding | 0.80 (0.19–2.44) | 0.721 | ||
RVAD requirement at 3 months | 1.98 (0.47–5.67) | 0.309 | ||
Inotropic support duration (days) | 1.02 (1.00–1.04) | 0.038 | ||
CVVHD requirement | 3.27 (0.90–9.66) | 0.070 | ||
Bacteremia during acute phase | 2.82 (1.33–6.21) | 0.003 | 2.36 (1.02–5.62) | 0.044 |
Laboratory data at 3 months | ||||
Hemoglobin (g/dL) | 0.75 (0.58–0.95) | 0.016 | 0.99 (0.75–1.29) | 0.918 |
WBC count (×103/μL) | 1.02 (0.84–1.18) | 0.860 | ||
C-reactive protein (mg/dL) | 1.30 (1.10–1.48) | 0.004 | 1.23 (1.01–1.44) | 0.042 |
Creatinine (mg/dL) | 1.24 (0.51–2.26) | 0.586 | ||
Total bilirubin (mg/dL) | 0.97 (0.52–1.02) | 0.559 | ||
Lactate dehydrogenase (×100 IU/L) | 1.26 (1.09–1.42) | 0.002 | 1.27 (1.09–1.46) | 0.004 |
Post-LVAD LVDd (mm) | 0.99 (0.96–1.01) | 0.244 |
*BMI and BSA at 3 months after LVAD implantation. Other abbreviations as in Tables 1–3.
We also evaluated the time relationship between the occurrence of bacteremia and hemorrhagic stroke. Findings of our analysis of the incidence of hemorrhagic stroke after the first bacteremic episode are shown in Figure 4A. Of 51 patients who developed bacteremia (7 who had their first bacteremic episode after stroke were excluded), hemorrhagic stroke prevalence after the first episode of bacteremia was 22%, 24% and 31% at 90 days, 6 months, and 1 year, respectively. There were no significant differences associated with bacterial species (S. aureus vs. non-S. aureus (Figure S1A), gram-positive coccus vs. non-gram-positive coccus (Figure S1B)). We also analyzed the incidence of hemorrhagic stroke according to recurrence of bacteremia in these 51 patients (Figure S2). In the analysis, although the difference did not reach statistical significance, patients who had recurrence of bacteremia had a slightly higher incidence of hemorrhagic stroke. Furthermore, 16 patients developed hemorrhagic stroke within 1 year of their last bacteremic episode. Of these, 11 (69%) developed hemorrhagic stroke during the first 90 days after the last bacteremic episode. To evaluate the time-dependent risk of hemorrhagic stroke after a bacteremic incident, we analyzed the transition of hazard rate of hemorrhagic stroke events in those 16 patients based on the interval from their last bacteremic episode (Figure 4B). The slope of the hazard ratio was the highest immediately after the onset of bacteremia, and the risk steadily decline to the normal level by 90 days after the last episode of bacteremia and flattened thereafter.
(A) Incidence of hemorrhagic stroke after the first episode of bacteremia. (B) Transition of 16 hemorrhagic stroke events that occurred within 1 year after the last episode of bacteremia, according to the interval from the last episode of bacteremia.
In the present study we found that (1) hemorrhagic stroke was more prevalent than ischemic stroke in our study patients with an LVAD, (2) bacteremia was an independent risk factor for hemorrhagic stroke in the chronic phase, while older age, but not bacteremia, was a risk factor for that in the acute phase, and (3) the risk of hemorrhagic stroke was high in the early phase after a bacteremic incident and then gradually decreased.
In the present study, though the prevalence of ischemic stroke was similar, hemorrhagic stroke was more prevalent than in other previous reports even in patients without bacteremia.3,4 One of reasons for this high prevalence of hemorrhagic stroke in our study population might be racial differences. In the previous reports that evaluated the prevalence and risk factors of stroke almost all studies comprised only Caucasian and/or African-American patients.4,7 As for racial differences, there has been only 1 study that compared the incidence of stroke between Caucasian and African-American LVAD patients,8 and none with solely Asian patients. Even in the general population, the incidence of cerebrovascular accidents and the ratio between hemorrhagic and ischemic stroke differ according to racial differences.9,10 Though it might be reasonable that the prevalence of stroke differs among the races, differences in vessel fragility and coagulation function among races need to be further evaluated in LVAD patients.
LVAD-related infection is one of the most prevalent complications and causes of death in patients receiving LVAD support.1 Several factors (e.g., driveline infection, pump pocket infection) can cause bacteremia. In the present study, a requirement for postoperative continuous venovenous hemodialysis was the independent risk factor for bacteremia, and patients with a low preoperative serum albumin level showed a trend of higher incidence of bacteremia. Toda et al reported that hemodialysis requirement, right heart failure, and longer ventilation were risk factors for bloodstream infection in LVAD patients.11 Their results, together with those from the present study, suggest that patients with a more severe preoperative condition (right heart failure, end-organ dysfunction, cachexia) are vulnerable to postoperative bacteremia.
In the present study, the risk factors for hemorrhagic stroke varied according to the postoperative period. There have been few studies that evaluated the risks of hemorrhagic stroke in patients after dividing the postoperative period into acute and chronic phases.12 An episode of bacteremia in the present patients was an independent risk factor for hemorrhagic stroke occurring not in the acute phase, but in the chronic phase. Previous reports have shown that a persistent bloodstream infection increases the risk of hemorrhagic stroke in LVAD patients.5,6,12,13 Infectious diseases accompanied by bloodstream infection (e.g., infective endocarditis) can frequently cause mycotic angiopathy and hemorrhagic cerebral complications.14–16 Therefore, it is reasonable to assume that systemic bacteremia from an LVAD-related infection (e.g., deep driveline infection, pump pocket infection) leads to cerebral mycotic angiopathy, which can occur as a devastating vascular rupture. In addition, the effect of non-physiological continuous pulseless flow on vascular fragility under bacteremic conditions should be further evaluated.
The present study analyzed the time relationship between systemic bacteremia and hemorrhagic stroke in LVAD patients. In a study of endocarditis, the risk of stroke, especially hemorrhagic stroke, was reported to be high in the first week of endocarditis, while persistent bacteremia was an independent risk factor for stroke.17,18 Although not all the findings from these endocarditis studies apply to bacteremia occurring in LVAD patients, our results suggested that extra attention should be given to cerebral complications, especially in the early period after an episode of bacteremia, and that bacteremia in LVAD patients must be treated immediately with prolonged, sufficient antibiotics therapy and/or surgical intervention when needed.
Study LimitationsFirst, this study was a retrospective study in a single center and the number of patients was inevitably limited. Second, 5 different types of LVAD were included, because the available devices varied during the study period. Although there were no statistical differences among the various LVADs used, the numbers of each device were further limited and the analysis might be under-powered. Finally, we did not completely distinguish between cerebral or cerebellar hemorrhage and hemorrhagic transformation of a cerebral or cerebellar infarction. When a transformed hemorrhage proceeds beyond the infarct area, accurate discrimination is nearly impossible.
In conclusion, hemorrhagic stroke events could be divided into those occurring in the acute and chronic phases. Bacteremia was shown to be an independent risk factor of hemorrhagic stroke in the chronic phase and the risk of hemorrhagic stroke was highest in the early period after a bacteremic incident. Further studies are warranted to establish the optimal precautions for bacteremia and hemorrhagic stroke in patients receiving LVAD support.
None of the authors have conflicts of interest to declare in regard to this study. Several of the clinical trial cases were subsidized by a Grant-in-Aid for Scientific Research from the Ministry of Health, Labor and Welfare of Japan.
Supplementary File 1
Figure S1. Incidence of hemorrhagic stroke according to bacterial species.
Figure S2. Incidence of hemorrhagic stroke according to the recurrence of bacteremia.
Table S1. LVAD stratified model analysis for hemorrhagic stroke
Please find supplementary file(s);
http://dx.doi.org/10.1253/circj.CJ-17-0541