Supplemental Digital Content
METHODS
A. Study design
This single-center prospective longitudinal cohort study with one-year follow-up (NCT02276417) included 239 surgical sepsis patients hospitalized at 48-bed surgical intensive care unit (ICU) in quaternary care hospital, University of Florida (UF) Health, between January 2015 and July 2017 (SDC Figure 1). The ethics approval was obtained from the University of Florida Institutional Review Board (201400611) and informed consent was obtained from each subject or their surrogate decision-maker.
Inclusion criteria were admission to the intensive care unit (ICU), age ≥18 years, and a clinical diagnosis of sepsis with subsequent initiation of a clinical decision support-directed sepsis management protocol by ICU team.1,2 Patients with a Do Not Resuscitate or Do Not Intubate order were eligible for enrollment if the patient or patient’s family were committed to aggressive management (e.g., initiation of hemodialysis) and escalation of care other than cardiopulmonary resuscitation and intubation. Patients with pre-admission end-stage renal disease, advanced liver or heart disease, systemic immunological disorders, those who were transplant recipients, immunosuppressed, or under cancer chemotherapy were excluded.1 During weekly meetings study investigators clinically adjudicated sepsis diagnoses using Sepsis-2 criteria.3 Anatomic site of inciting infection was clinically adjudicated by clinical acute care surgeons (JS, TL, and SB). Abdominal sepsis included any primary infection arising from the gastrointestinal/hepatobiliary tract or secondary to a recent complication of gastrointestinal/hepatobiliary surgery (leak, abscess, and fistula). Pulmonary infections included pneumonia and empyema. Skin/soft tissue infections included primary soft tissue infections and surgical site infections (including superficial incisional, deep incisional, and organ/space infections). Genitourinary infections included infection arising from the urinary tract or the male
or female genital tract tracts. Vascular infections included mycotic aneurysm, septic thrombophlebitis, infected prosthetic grafts or a central-line associated bloodstream infection.
Blood and urine samples were collected to characterize biological correlates of endothelial function, inflammation, and immunosuppression, within twelve hours of sepsis onset and at predefined time points up to fourteen days; 1 day, 4 days, 7 days, and 14 days (SDC Table 1). Clinical assessments of functional and renal status were performed in-person and with telephone visits at three, six, and twelve months after enrollment. 4,5
B. Assessment of kidney function
Three independent nephrologists (AB, MS, RM) clinically adjudicated all acute kidney injury (AKI) episodes using available clinical information and according to Kidney Disease:
Improving Global Outcomes criteria (0.3 mg/dl increase in serum creatinine within 48 hours or 50% increase from baseline within seven days or decrease in urine output to less than 0.5 ml/kg/hr for six hours).6 The maximum AKI stage was determined based on the ratio between peak and reference serum creatinine and need for renal replacement therapy (RRT). Based on duration of AKI episode and evidence of renal recovery7, we defined clinical trajectories of rapidly reversed and persistent AKI with and without renal recovery at discharge. Persistent AKI was defined as an episode of AKI lasting for at least 48 hours after the onset. Any episode of shorter duration was characterized as rapidly reversed AKI. Renal recovery was adjudicated for each episode based on normalization of AKI criteria at the time of hospital discharge.6,7 AKI that developed within 48 hours of sepsis onset was considered as early AKI. Recurrent AKI was defined as more than one AKI episode during index hospitalization that were separated by at least 48 hours with no evidence of AKI.
To determine reference creatinine, we used previously validated modification of the NHS England alert algorithm. 8-10 For patients with available preadmission measurements (n=175) reference value was defined as either the lowest in the last 7 days or a median of values from
the preceding 8 to 365 days depending on availability of previous results. For patients with no available preadmission measurements and no history of chronic kidney disease (CKD) we used the lowest of admission creatinine and estimated baseline creatinine using the Modification of Diet in Renal Disease Study equation assuming that baseline estimated glomerular filtration rate (eGFR) is 75 ml/min per 1.73 m2 (n=64).6,11,12 For patients with known history of CKD and no available preadmission measurements we used lowest creatinine value on admission day (n=5).
After first seven days of hospitalization, minimum serum creatinine measurements in preceding 7 days was used as the reference creatinine.
Reference creatinine was used to estimate preadmission reference glomerular filtration rate using Chronic Kidney Disease Epidemiology Collaboration equation.13 Chronic kidney disease was determined from medical histories obtained prospectively at the time of enrollment and from a validated combination of International Classification of Diseases codes from
electronic health records.14 Chronic kidney disease stages were determined based on reference eGFR according to guidelines.15,16
For each patient, we calculated daily kinetic GFR using the estimate of creatinine production rate and percent change in creatinine over time.17 We used creatinine values separated by at least twelve hours to estimate creatinine production rate and provides information on percent change in creatinine over time. We applied following formula:
KeGFR=SSPCr×CrCl
MeanPCr ×
(
1−ΔTime(h)24× MaxΔ P× Δ PCr Cr/Day)
17using creatinine production rate that can be expressed as the product of initial serum creatinine ( SSPCr¿ and creatinine clearance ( CrCl ), average of consecutive creatinine measurements ( MeanPCr¿ separated by ΔTime(h) hours with difference Δ PCr .
MaxΔ PCr/Day refers to the maximal change in creatinine which is taken as 1.5.
Cumulative fluid balance on each day was calculated as total fluid balance from hospital admission divided by admission weight in kilograms using the equation
Cumulative fluid balance(%)=(cumulative daily fluid input−output)(L)❑
Hospital admission weight(kg) ×100 . 18
Microalbumin to creatinine ratio was measured at the sepsis onset and at 12-month follow up. Serum creatinine measured at 12-month follow up was used to calculate follow-up eGFR using Chronic Kidney Disease Epidemiology Collaboration formula13 and calculate change in eGFR. For patients who were dialysis dependent eGFR was estimated as 0. For 25 patients who died between discharge and 12-months follow-up we used the last measured renal function. Change in eGFR at 12 months compared to preadmission eGFR were reported for 140 patients including those who came for follow up and those who died after discharge before one- year follow-up time.
C. Biological correlates, functional assessment and clinical outcomes
Chronic disease burden was characterized by Charlson-Deyo comorbidity index scores.
19 Severity of illness was characterized by Acute Physiology, Age and Chronic Health Evaluation II (APACHE II) scores within the first 24 hours of sepsis onset and by daily Sequential Organ Failure Assessment Score (SOFA). 1 Organ dysfunction was defined as SOFA score two or more, which occurs with any of the following criteria: Glasgow coma scale <13, vasopressor or inotrope administration (including epinephrine, norepinephrine, dopamine, dobutamine,
phenylephrine, and vasopressin), mechanical ventilation, ratio of partial pressure of arterial oxygen and fraction of inspired oxygen (PAO2/FiO2)< 300 mmHg, serum creatinine > 2 mg/dl, platelets < 100,000/µl, or bilirubin > 2 mg/dl. Lowest and highest GCS within 24 hours of sepsis onset were accrued excluding time periods on sedation and imputing missing Glasgow coma scale values with 15 as a sensitivity analysis. Temperature measured from oral and axillary sites
were adjusted to the rectal site by adding 0.5 and 1 degrees Celsius, respectively. Values outside of ranges (20, 300) for systolic, (5, 225) for diastolic, and (10, 250) mm Hg for mean arterial blood pressure were removed from the dataset as well as systolic blood pressure values if it is less than diastolic blood pressure + 5 mm Hg. 20 Missing mean arterial blood pressure values were imputed with the sum of the 2/3 of diastolic and 1/3 of systolic blood pressure.
Missing fraction of inspired oxygen (FiO2) values were imputed based on respiratory device and oxygen flow rates. 21 Blood products considered were red blood cells, plasma, platelets, and cryoprecipitate. We determined longitudinal changes in ten serum biomarkers reflective of the host response for sepsis broadly categorized under endothelial function, inflammation,
immunosuppression, and metabolism (SDC Table1). Follow-up assessments were performed 3, 6, and 12 months after sepsis onset with in-person visits conducted at the University of Florida’s Institute of Aging or the patient’s home to quantify functional status, and telephone calls were made if it was not possible to see the subject in person. The five-point Zubrod Scale was used to measure and compare physical performance status (0 = fully functional/active and able to carry on all pre-disease activities without restrictions, 1 = restricted in physically strenuous activity but completely ambulatory, 2 = ambulatory and capable of all self-care but unable to perform any work activities, spending less than 50% in bed during the day, 3 = capable of only limited self-care, spending greater than 50% of the day in bed, 4 = bedbound, incapable of self- care, 5 = deceased). 4,5
The primary clinical outcome were thirty-day mortality and one-year survival. The primary renal outcomes were major adverse kidney events at 30 and 365 days after sepsis onset , defined as composite outcome of death, dependency on renal replacement therapy, or decrease in an eGFR <60 ml/min/1.72m2.
Other exploratory outcomes included hospital-free, ICU-free and mechanical ventilation- free days within 28 days of sepsis onset and functional recovery at one-year. Exact dates and times were used to calculate the hospital length of stay, ICU length of stay, and duration of
mechanical ventilation. Hospital-free, ICU-free, mechanical ventilation-free, and organ dysfunction free-days within 28 days of sepsis onset were calculated by subtracting the number of days for each of outcome from the lesser of 28 days or the number of days between sepsis onset and death. 22 The Social Security Death Index database was used to confirm death dates and obtain death dates for subjects who were lost to follow-up.
D. Biomarker analyses
Blood and urine samples were analyzed for biomarkers of endothelial function, inflammation, and immunosuppression. The summary of preselected biomarkers’ function and values determined among healthy controls is reported in SDC Table 1. Serum levels of vascular endothelial growth factor and plasma levels of soluble programmed death-ligand 1, glucagon-like peptide 1, erythropoietin, soluble vascular endothelial growth factor receptor-1, and angiopoietin 2 were determined by ELISA (R&D Systems, Minneapolis, MI, USA). Plasma levels of tumor necrosis factor, interleukin 6, monocyte chemoattractant protein 1, interferon gamma-induced protein 10, and interleukin 8 were measured using Luminex multiplex kits (MILLIPLEX multiplex assay, EMD Millipore Corp. Billerica, MA, USA). Complete blood counts with leukocyte differentials, automated urinalysis and microalbumin/creatinine ratio were measured by the Clinical and Diagnostic Laboratories at UF Health.23 For all ELISA and multiplex analyses, an identical internal control sample was used across all kits to normalize the data.
E. Statistical analysis
The sample size offered 80% power to detect an 8% difference in thirty-day mortality and two-fold increase in the hazard ratio for one-year mortality between patients with and without AKI, assuming AKI prevalence of 62% and thirty-day and one-year mortality in the no AKI group of 1% and 10%, respectively (Type 1 error < 0.05).
Differences between groups were analyzed using the Kruskal-Wallis test or analysis of variance for continuous variables and the chi-square or Fisher’s exact test for discrete variables as appropriate, adjusting for multiple comparisons between trajectory groups using Steel- Dwass-Critchlow-Fligner multiple comparison procedure for nonparametric comparison of continuous variables and Bonferroni method for others. Overall survival of each trajectory group was evaluated using log-rank Kaplan–Meier methods. Propensity score based inverse weighting was used to plot adjusted Kaplan Meier curves where propensity of being in a trajectory group was calculated using multinomial logistic model that included expert-selected variables which are patient demographics (age, gender, African-American ethnicity), Charlson comorbidity score, and sepsis shock status and non-renal SOFA score on sepsis onset. Cox proportional-hazards regression was used to assess associations between AKI trajectories and one-year mortality while controlling for total ICU length of stay in addition to the same variables included in the propensity score models. There were no missing values for input and outcome variables used for the models. Survival models were started at the time of sepsis protocol initiation and followed up to 15 months. Using scaled Schoenfeld residuals we confirmed that the proportional hazards assumption was satisfied for all variables in the model. Results were reported as unadjusted and adjusted hazards ratios with 95% confidence intervals. Model discrimination was assessed using Harrell’s concordance index. Cumulative net fluid balance and kinetic GFR values over time were shown by locally estimated scatterplot smoothers calculated utilizing local regression using R function “loess”. 24 A p-value ≤ 0.05 was considered statistically significant. Statistical analyses were performed with SAS (v.9.4, Cary, NC), R 3.5.3, and Python 3.6.6 software.
References
