Participants
This study was conducted at a single center, specifically the intensive care unit (ICU) of the First Affiliated Hospital of Zhejiang University School of Medicine. We enrolled patients who received ECMO treatment from August 2020 to December 2023 in this study. These patients included patients who received VA ECMO and VV ECMO support. Exclusion criteria consisted of 1) Age below 18 years old; 2) History of cerebral disorders such as cerebrovascular accident, intracranial tumor, central nervous system immune disease, dementia, etc.; 3) Inability to undergo cranial magnetic resonance imaging examination, 4)modes was converted during ECMO operation, and 5) died within 90 days after ECMO withdrawal.
Clinical data
The clinical data for this study were meticulously gathered through an inpatient data management system, capturing a wide array of demographic information. Medical histories include cardiovascular risk factors such as hypertension, hyperlipidemia, diabetes mellitus, smoking, recently described medication, and previous stroke. Blood coagulation parameters during the ECMO operation were recorded. This comprehensive assessment included the Activated Clotting Time (ACT), Prothrombin Time (PT), D-Dimer levels, activated Partial Thromboplastin Time (aPTT), and Fibrinogen (Fib) levels. In addition to these parameters, the study also incorporated several other critical markers of patient status and response during ECMO treatment. This encompassed the highest Sequential Organ Failure Assessment (SOFA) score, the maximum lactate level (Lac) recorded, the lowest hemoglobin (Hb) level observed, the worst Glasgow Coma Scale (GCS) score, and the timing of GCS Motor (GCS-M) ≥ 5 points.
Patient management under ECMO
Blood gas analysis was conducted on all patients within one hour before and after ECMO initiation, with subsequent reviews every four hours during the ECMO operation. All patients received heparin anticoagulant therapy, administered as a loading dose before ECMO. Heparinization was adjusted at least once daily based on activated partial thromboplastin time (aPTT) values (VV-ECMO: 60-80s; VA-ECMO: 80-100s) and clinical tolerance. Heparin was discontinued in cases of bleeding and reintroduced when bleeding was controlled. Bleeding leading to discontinuation of heparin refers to any clinical bleeding (from the ECMO site, central or arterial lines, tracheal secretions, ear, nose, or throat), with or without hemodynamic effects and/or a decrease in hemoglobin levels as determined by the attending physician. Membrane function and loop pressure changes were assessed every 24 hours. During the ECMO procedure, all patients were routinely administered sedatives and analgesics, while monitoring the depth of sedation using BIS. If conditions allow, it is recommended to discontinue sedative and analgesic medications every 24 hours. The GCS score was assessed by specialized ICU physicians or neurologists based on a BIS score ≥ 80 or if the BIS score did not increase after discontinuation of anesthetic drugs.
Outcome Parameters
The primary objective was to assess the impact of different intensities of anticoagulation parameters on the incidence and severity of cerebral microbleeds (CMBs) in patients undergoing ECMO. The secondary objective includes evaluating the association between varying degrees of CMBs and modified Rankin Scale (mRS) scores at 90 days, cerebral performance category (CPC) scores, and an additional association between the degree of recovery of Glasgow Coma Scale-Motor (GCS-M) score within the first two weeks and the severity of CMBs on MRI.
MR Imaging
All eligible subjects were scanned on a GE HDx 3.0T magnetic resonance scanner using a combined head and spine coil. Participants underwent conventional MRI plain scan and swan sequence scan. Conventional MRI plain scan sequence and parameters: FSE T1-fluid attenuated inversion recovery (FLAIR) sequence, TR 1930ms, TE 22ms or RT 2500ms, TE 9ms, T 720ms in axial and sagittal planes, Matrix 320 × 224, excitation times 1; propeller sequence T2WI, TR 6400ms, TE 134ms, matrix 320 × 320, excitation times 1.5; FSE axial T2 FLAIR sequence, TR 8502ms, TE 162ms, TI 2100ms, matrix 256 × 192, excitation times 1; DWI sequence, TR 4800ms, TE 80ms, matrix 96 × 130, excitation number 2; In the horizontal plane, the field of view was 24cm × 24cm, the thickness was 6mm, and the interval was 1mm. The sagittal field of view was 22cm × 22cm with 5mm slice thickness and 1mm interval.
The STandards for Reporting Vascular Changes on nEuroimaging (STRIVE) consensus criteria were used to rate MRI markers of cerebral small vessel disease including lacunes, cerebral microbleeds (CMBs), white matter hyperintensity (WMH), and perivascular spaces (PVS)7. On FLAIR and T2 scans, lacune lesions most showed a hyperintense rim and a slight increase in signal on FLAIR; they were described as rounded or ovoid lesions involving the subcortical regions, 3–15 mm in diameter of CSF signal intensity. WMH was assessed on FLAIR images using the Fazekas scale8. In susceptibility-weighted imaging (SWI), CMBs appeared as homogenous, rounded hypointense lesions ranging in size from 2–10 mm. CMB burden was categorized as low (< 10 CMB), moderate (10–30 CMB), and high (> 30 CMB) as previously reported6. Figure 1 shows the stratification of CMB burden.
Statistical analysis
Continuous variables with normal distribution were expressed as mean ± standard deviation (SD), while skewed distribution was shown as medians and interquartile ranges. Categorical variables are presented as frequencies and percentages (%).
Multivariate regression analysis was used to explore the relationship between CMB, clinical outcomes, and blood parameters controlling for key covariates: age, gender, hypertension, diabetes mellitus, hyperlipidemia, smoking, and ECMO operation time.
Mediation analysis was performed to assess whether the association between blood parameters and clinical outcomes was mediated by CMB. Nonparametric bootstrapping (B = 1000) was used to compute 95% confidence intervals (95% CI) for the total effect, mediation effects, and direct effects. The covariates for multivariate regression analysis were set as age, gender, hypertension, diabetes mellitus, hyperlipidemia, smoking, and ECMO operation time. Statistical analysis and plotting were conducted in R version 4.2.3 (gee package for GEE, mediation package for mediation analysis), and p < 0.05 was considered significant.