Study patients and design
This multicenter retrospective study was performed using data from three ICUs located at Fujian Medical University Union Hospital and Fujian Provincial Hospital with a total of 85 beds from January 1, 2013 to December 31, 2019. All participants underwent transthoracic echocardiography within 24 hours of admission to identify the presence or absence of LV dysfunction. The exclusion criteria included: younger than 18 years of age, moderate-to-severe valvular heart disease, history of end-stage renal disease or hemodialysis, postrenal causes of renal injury, cardiopulmonary resuscitation before ICU admission, intoxication, cirrhosis, rhabdomyolysis, active malignancy, connective tissue diseases, pregnancy, expected survival less than 24 hours, normal LV function, poor echocardiographic image quality.
All patients included in this study were managed with CRRT. Some hemodynamically unstable patients receiving CRRT did not have septic AKI before CRRT. Patients were divided into one of two groups according to the baseline AKI status: the early group (no AKI) or the control group (with AKI). In the early group, early initiation of CRRT was performed in the absence of AKI, though AKI could occur thereafter. The control group received CRRT when AKI was presented. CRRT was discontinued when hemodynamics or renal function was improved. Clinical outcomes included all-cause ICU mortality and length of ICU stay.
Data collection
Data concerning demographic and clinical information (primary diagnosis and comorbidities), physiological parameters (hemodynamic data and vasoactive medications), transthoracic echocardiographic parameters, laboratory results, and the use of invasive mechanical ventilation were extracted from electronic medical records by trained medical staff. Information of the CRRT was reviewed. The urine output (UO) and serum creatinine levels were obtained to verify the presence of AKI. The corresponding glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease (MDRD) equation[12]. Baseline disease severity was assessed by the Sequential Organ Failure Assessment (SOFA) score and Acute Physiology and Chronic Health Evaluation (APACHE) II score.
Definitions: Severe sepsis, septic shock, and septic AKI
Severe sepsis was defined as sepsis related to organ dysfunction, hypoperfusion, or hypotension[13]. A lactate level ≥ 2.3 mmol/L (22.1 mg/dl) was considered indicative of hypoperfusion. Hypotension was defined as systolic blood pressure ≤ 90 mmHg or a decrease of 40 mmHg below baseline, organ dysfunction as SOFA score ≥ 2[13]. Septic shock was defined as sepsis-induced persistent hypotension requiring vasopressor therapy to maintain a mean arterial pressure (MAP) of ≥ 65 mmHg or a lactate level ≥ 2.3 mmol/L (22.1 mg/dl) after adequate fluid resuscitation[14, 15]. Septic AKI was defined as the simultaneous presence of sepsis criteria[14] and the consensus criteria for AKI according to the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines[16]. The baseline creatinine value was either obtained from clinical files within seven to 365 days previous to admission or the minimum inpatient values during the first seven days of admission[7]. AKI was defined as meeting one of the following criteria: an increase in creatinine by ≥ 0.3 mg/dl within 48 hours; an increase in creatinine to ≥ 1.5 times baseline within the previous seven days; or urine output ≤ 0.5 ml/kg/h for 6 hours.
Transthoracic echocardiographic examination
All echocardiograms were assessed by a professional cardiologist. Two-dimensional, M-mode, and Doppler data were used to obtain parameters from parasternal long- and short-axis views; apical four-chamber, two-chamber, and long-axis views; and subcostal views. Data on early diastolic velocity of mitral inflow (E), early diastolic mitral annular velocity (e'), late diastolic velocity of mitral inflow (A), E/e' ratio, and E/A ratio were collected. According to the American Society of Echocardiography 2009 guidelines[17] and the simplified definition suggested by Lanspa et al.[18], LVSD was defined as LV ejection fraction (LVEF) < 50%, and LV diastolic function was classified into four grades (normal and grades I, II, and III).
CRRT settings
CRRT was performed in either continuous veno-venous hemofiltration (CVVH) or continuous veno-venous hemodiafiltration (CVVHDF) through the femoral or internal jugular veins at the discretion of attending physicians. PRISMAFLEX and AQUARIUS hemofiltration systems were used with the addition of bicarbonate or potassium if necessary. The dialysate rate, replacement fluid rate, and ultrafiltration rate were adjusted according to patients’ diagnoses, hemodynamic parameters, and fluid overload. AN69 membranes or RENAFLO hemofilters were used and blood flow rates were kept between 100 and 200 mL/min during the procedure.
Statistical analyses
Continuous variables were expressed as the mean ± standard deviation for normally distributed data and differences between groups were determined using a two independent samples t-test. Data without normal distribution were expressed as the median (interquartile range, P25, and P75) and two groups were compared using the Mann-Whitney U test. Categorical variables were presented as counts (percentages) and compared using Pearson’s chi-square test or Fisher’s exact test.
Propensity score weighting (PSW) was applied to balance the baseline characteristics between groups. Firstly, the baseline characteristics were compared between the two groups. Secondly, logistic regression analysis was used to evaluate the probability of treatment with early CRRT or not. With the treatment allocation as dependent variables, and the factors with p values < 0.10 between the two groups at admission were taken as the candidate independent variables. The logistic regression model was constructed to calculate the individual propensity score. Thirdly, patients were weighted by the stabilized inverse probability of treatment weighting (sIPTW) and the weighted baseline characteristics were tested again. Clinical outcomes including ICU mortality and length of ICU stay were compared between two groups by chi square analysis and an independent samples t-test before and after weighting.
The risk factors associated with ICU mortality were further analyzed, and the impact of early CRRT on mortality was evaluated. With mortality as the dependent variable, and baseline clinical characteristics at admission and at the start of CRRT as independent variables, the weighted univariate logistic regression analyses were conducted separately. According to the results of univariate regression analysis, the factors with p < 0.05 were selected as the candidate independent variable to construct the multivariate weighted logistic regression model. The step-by-step method was used to screen the variables. Effect size was presented as the odds ratio (OR) with the corresponding 95% confidence interval (CI). All data were analyzed using R 4.0.2 software, and p < 0.05 was considered statistically significant.