Design
We conducted, between 06/01/2011 and 07/31/2018, a single center retrospective cohort study on consecutive IE patients admitted to the medical ICU of a tertiary referral center for cardiac surgery and IE treatment, in Paris, France. In this center, patients with IE are treated by a multidisciplinary IE team including cardiologists, infectious disease specialists, microbiologists, cardiac surgeons, neurologists, radiologists, anesthesiologists and intensive care specialists.
Patients
Patients were screened via a computer search in the local discharge database according to the ICD-10 (code I33 (acute and subacute endocarditis) and/or I38 (endocarditis, valve unspecified)). We included all consecutive patients ≥ 18 years old diagnosed with a definite, active, severe, left-sided IE requiring ICU admission. Definite IE was defined according to the modified Duke criteria [11]. IE was defined as active if the patient was admitted to the ICU before or within the first 30 days of antimicrobial treatment. Severe IE was defined as a SOFA score ≥ 2, indicating the presence of an organ dysfunction [16]. Left-sided IE was defined as IE involving mitral and/or aortic valves. Cases of both left- and right-sided IE or with left-sided IE associated with infection of a cardiac implantable electronic device (CIED) were also considered as left-sided IE. Exclusion criteria were: (1) patients transferred to the ICU after cardiac surgery for IE, (2) ICU-acquired IE, (3) patients already included in the study for a previous IE and (4) patients with no available baseline brain CT. Baseline brain CT was defined as the first brain CT performed after ICU admission or, if no brain CT had been done in the ICU, as the last brain CT before ICU admission. CT-scans performed after IE surgical treatment or > 10 days before ICU admission were not considered as baseline brain CT.
Initial workup
All IE patients received an initial standardized diagnostic workup including clinical evaluation, echocardiography (trans-thoracic echocardiography (TTE) and trans-esophageal echocardiography (TOE)), routine biological workup and a bacteriological workup including at least two blood sample cultures. All TTE/TOE were conducted by a senior cardiologist from the cardiac surgery department. The vegetation length was measured in various planes during the first echocardiogram and on follow-up when available, and the maximal length was used for analysis. Other explorations including coronary angiography, CT scan of thorax, abdomen and pelvis (CT-TAP) and brain vascular imaging were performed when deemed necessary.
Ethics
The local ethics committee (Comité local d’éthique HUPSSD-Avicenne) approved the study (agreement CLEA-2021-191).
Data collection
Each patient’s condition at ICU admission was assessed using the Sequential Organ Failure Assessment (SOFA) [16], and the GCS scores [17], both prospectively calculated. If the patient was sedated at ICU admission, the last GCS available before sedation was recorded and used in the analysis. If the initial brain-CT scan revealed a stroke, the clinical severity was assessed using the National Institute of Health Stroke Scale [18] (NIHSS). When necessary, the NIHSS was assessed retrospectively by a neurologist (TR) using a validated approach [19]. Neurological symptoms were assessed retrospectively by a neurologist (TR), based on the systematic initial clinical examination conducted by a senior intensivist. Immunodepression was defined as human immunodeficiency virus infection, malignancy, long-term use of corticosteroids or other immunosuppressants.
IE
Community-acquired IE was defined as IE diagnosed within 48 h of hospital admission, for a patient who did not satisfy the criteria for health care-associated infection. Healthcare-associated IE was classified as nosocomial or non-nosocomial in accordance with current definitions [20]. IE diagnosis date was defined as the day on which the first positive blood culture was obtained or, in case of blood culture-negative IE, as the date of the first echocardiography revealing findings consistent with IE.
Surgery
Indication for surgery was assessed throughout the ICU stay according to European guidelines [11] and further classified as emergency (within 24 h), urgent (within 7 days) or non-urgent (surgery can be postponed to allow 1 or 2 weeks of antimicrobial therapy).
Neuroimaging
The first report of the baseline brain CT scan, carried out by a senior radiologist, was collected. Images from this CT-scan were re-analyzed by a neurologist (TR), using a standardized method and blinded to the first report of the imaging, clinical data, outcome and any other brain imaging undergone by the patient. Findings from the radiology report were then compared to the second interpretation and, in case of discrepancies, a third interpretation was made by another neurologist (PJ), also blinded to all relevant data and to both previous interpretations.
The presence of brain lesions was assessed on the baseline brain CT scan, according to an a priori defined grid: ischemic stroke (IS) including topography and volume (evaluated using the semi-quantitative ASPECT and PC-ASPECT scales [21, 22]), hemorrhagic transformation (HT) (classified according the ECASS-2 classification [23]), intraparenchymal hemorrhage (IPH), subarachnoid hemorrhage (SAH), subdural hemorrhage (SDH), extradural hemorrhage (EDH), abscess, empyema and mycotic aneurysm.
Baseline brain CT-scans were classified in five mutually exclusive categories: normal, moderate-to-severe ischemic stroke, minor ischemic stroke, cerebral hemorrhage and other infectious complications.
Moderate-to-severe ischemic stroke was defined (adapted from [6]) as an ischemic stroke involving the middle cerebral artery (MCA) territory with an ASPECT score ≤ 7 and/or involving the posterior cerebral artery territory with a PC-ASPECT score ≤ 7 and/or a multiterritorial ischemic stroke with at least two lesions ≥ 15 mm involving at least two distinct vascular territories. All ischemic strokes not meeting these criteria were classified as “minor ischemic stroke”. Cerebral hemorrhages included IPH, SAH, HT (except minor non-confluent petechiae [23]), SDH and EDH. Other infectious complications included cerebral abscess, empyema and diffuse cerebral edema (adapted from [6]).
Patients with several different radiological lesions were classified as follows: (1) patients with a normal baseline CT were classified in the “normal CT” category; (2) patients with moderate-to-severe ischemic stroke and any other radiological lesions were classified in the “moderate-to-severe ischemic stroke” category; (3) patients with a cerebral hemorrhage and either a minor ischemic stroke or another infectious complication were classified in the “cerebral hemorrhage” category; (4) patients with a minor ischemic stroke and another infectious complication were classified in the “minor ischemic stroke” category (Additional file 5: Figure S1 and Additional file 6: Figure S2).
Endpoints
The primary endpoint was the proportion of patients with a favorable functional outcome, 1 year after ICU admission. Outcome was assessed using the modified Rankin Scale (mRS, ranging from 0 to 6 with a score of 0 indicating no disability, higher scores indicating more severe disability and 6 indicating death [24]). Favorable functional outcome was defined as a score on the mRS between 0 (asymptomatic) and 3 (moderate disability). The mRS was assessed using a standardized validated questionnaire [25], by reviewing the medical chart or, when the available data were insufficient, by contacting the patient’s treating physician. Secondary endpoint was the risk of severe post-operative neurological complications (defined as a new symptomatic neurological event with a new abnormality on any brain imaging and/or a new hospitalization or ICU admission due to a neurological event and/or death due to a neurological event) within 90 days after surgery.
Causes of death were classified as “neurological”, “heart failure or cardiogenic shock”, “sepsis or septic shock” or “other” by two neurointensivists (TR, RS) through analysis of medical charts.
Due to an expected very small number, patients with a baseline CT-scan classified as “other infectious complications” were excluded from outcome analysis.
Statistical analysis
Data are expressed as counts and frequencies for categorical variables and medians [interquartile range] for quantitative variables. Univariate comparisons between subgroups were performed using χ2 test or Fisher exact test for categorical variables and Mann–Whitney test for continuous variables.
To identify independent predictors of unfavorable outcome at one year, we used a multivariable logistic regression model using a backward selection procedure among the following variables: baseline brain CT categories and all baseline characteristics associated with the outcome of interest in univariate analysis (p < 0.1, Additional file 1: Table S1). To avoid collinearity, when both a score including several components (such as the Charlson comorbidity index or the non-neurological SOFA) and one of its individual components were associated with the outcome of interest, only the multi-component score was included in the multivariable analysis. Log linearity of continuous variables was tested and, if necessary, variables were categorized according to clinically relevant cutoffs or at median values. Collinearity between variables and 2-by-2 interactions were tested. A separate regression analysis was performed for each category of baseline brain CT. The magnitude of association with the outcome was expressed as an odds ratio (OR) and 95% confidence intervals (95% CI). Missing data were handled with pairwise deletion method. All tests were two-sided, and p-value < 0.05 was considered statistically significant. All analyses were performed using R, Version 3.5.2 (R Project for Statistical Computing, https://www.r-project.org).