A combination of SOFA score and biomarkers gives a better prediction of septic AKI and in-hospital mortality in critically ill surgical patients: a pilot study

This study was a prospective cohort study performed in a surgical intensive care unit (ICU) of a 3500-bed tertiary center in Taiwan. The surgical ICU is a 10-bed closed unit managed by a surgeon who is also certified to care for ICU patients. The study was approved by the Institutional Review Board of Chang Gung Memorial Hospital (IRB103-2722A3) and conducted according to the guidelines established by the Declaration of Helsinki. Written informed consent was obtained from all participants. The admission criteria to our surgical ICU include the following: (1) major postoperative complications requiring further invasive intervention; (2) complicated gastrointestinal disorders such as life-threatening gastrointestinal bleeding, fulminant hepatic failure, or severe pancreatitis; (3) hepatic failure complicated with multi-organ dysfunction related to liver transplantation; (4) acute postoperative respiratory distress; (5) cardiovascular instability such as shock of any cause; and (6) other clinical presentations deemed acceptable for ICU admission by the attending physician. We have consecutively recruited all patients admitted to the surgical ICU for 7 months spanning from November 2014 to June 2015. On ICU admission, the patients were initially excluded if they were less than 20 years of age and had a history of chronic kidney disease (CKD) for more than 3 months, inflammatory bowel disease, renal transplantation, or a need for a routine dialysis program. Applying the Sepsis-3 criteria or the identification of an infection site and an increase of greater than or equal to two points in the Sequential Organ Failure Assessment (SOFA) score [1], the patients were allocated into a non-septic group and a septic group. A baseline SOFA score of 0 was assumed for all patients. Among the septic group, some patients were further excluded if no consent form was obtained from the patient or next-of-kin, blood/urine specimens were not collected within the first 24 h of the ICU admission, or patients were transferred to our ICU with the initial treatments started in another ICU. The non-septic group was defined as those patients who had no clinical or laboratory evidence of infection. As this was a pilot study, it was predetermined that patients would be assigned to the septic or non-septic groups in a 2:1 ratio.

Upon admission, patients’ demographic information, comorbidities, type of surgery, use of vasopressors, the Acute Physiology and Chronic Health Evaluation (APACHE II), and the Sequential Organ Failure Assessment (SOFA) scores were recorded. SOFA was calculated sequentially based on the worst value of the parameters including partial pressure of oxygen (PaO₂), fraction of inspired oxygen (FiO₂), platelet count, bilirubin, mean arterial blood pressure (MAP), Glasgow Coma Scale (GCS), creatinine, and urine output over the preceding 24 h. The status of AKI was also documented and defined according to the risk, injury, failure, loss of kidney function, and end-stage kidney disease (RIFLE) criteria [8], as follows: a percentage decrease of > 25% in the glomerular filtration rate (GFR) or an increase of ≥ 1.5 times in serum creatinine (SCr) level is defined as risk; a > 50% decrease in GFR or ≥ 2 times increase in SCr is defined as injury; a > 75% decrease in GFR or ≥ 3 times increase in SCr is defined as failure; persistent acute renal failure more than 4 weeks is defined as loss; and complete loss of renal function for more than 3 months is considered end-stage renal disease (ESRD). The patients were followed throughout their ICU stay. CKD was defined as a GFR lower than 60 ml/min/1.73m² using the baseline creatinine and the CKD Epidemiology Collaboration equation [16].

Blood and urine samples were obtained as soon as possible after patients were admitted to the ICU. Blood samples were centrifuged at 1500g for 10 min, while urine samples were centrifuged at 500g for 10 min; both were aliquoted and stored at − 80 °C for batch analysis. Serum and urinary gelatinase-associated lipocalin (NGAL), calprotectin, KIM-1, cystatin C, and GDF-15 were measured using an enzyme-linked immunosorbent assay (ELISA) kit (DuoSet ELISA, R&D Systems; Minneapolis, MN, USA) [17‐23]. The dilution ratios for NGAL, calprotectin, KIM-1, cystatin C, and GDF-15 were 1:100, 1:1, 1:100, 1:400, and 1:100, respectively. Albumin, creatinine, procalcitonin, lactate, and C-reactive protein (CRP) were analyzed using enzymatic methods on an automated chemical analyzer in the central laboratory of Chang Gung Memorial Hospital.

The continuous variable data were tested for normality distribution using Kolmogorov-Smirnov and Shapiro-Wilk tests and presented as the mean ± standard error of mean (SEM). The independent sample t test and the Mann–Whitney U test were used for comparison of the study groups. Categorical variables were compared using Fisher’s test or Pearson’s chi-square test and presented as absolute frequency and percentages. Receiver operating characteristic (ROC) curve analysis was performed to determine the performance of biomarker concentration in the prediction of septic AKI and in-hospital mortality. Data analysis was performed using SPSS version 13 (SPSS Inc., Chicago, IL, USA). P values less than 0.05 were considered statistically significant.

Over the 7-month study period, a total of 315 patients were admitted to the surgical ICU, and after applying exclusion criteria, 83 patients were allocated to the septic group, as depicted in Fig. 1. Of the 83 patients with sepsis, 61 were further excluded from the study due to a refusal to participate (N = 42), an absence of next-of-kin to consent (N = 2), a failure to collect blood/urine specimens within 24 h of ICU admission (N = 13), and a transfer from another ICU where initial treatments had been started (N = 4). The baseline demographics and clinical characteristics of the 33 patients are shown in Tables 1 and 2. A majority of the patients were males (60.6%) and over 65 years of age (63.6%), with a mean age of 66.3 years. Among these 33 patients, 22 had a diagnosis of sepsis with varying degrees of AKI, while the remaining 11 were free of sepsis. In 15 (45.6%) patients, intra-abdominal infection was the sepsis etiology, 2 patients (6.1%) had pneumonia, 2 (6.1%) had Fournier’s gangrene identified as soft tissue infection, 1 (3.0%) had a urinary tract infection, while 2 (6.1%) had sepsis of unknown origin. Of the 16 patients finally diagnosed with AKI, 14 (42.4%) had a status of R, I, or F, according to the RIFLE criteria, while the other two had a status of L. As depicted in Tables 1 and 2, although 60% of the patients were not on vasopressors, 84.8% of patients required the assistance of mechanical ventilation, with an average of 8.9 ± 20.3 days before successful endotracheal extubation. The length of hospital stay ranged from 2 to 114 days (mean 39.1 ± 25.1 days), with a mean ICU stay of 12.6 ± 25.3 days.

Serum and urinary levels of calprotectin, NGAL, and cystatin C, in addition to the SOFA and APACHE II scores, were measured in septic AKI patients and represented as scatter dot plots in Fig. 2. Serum and urinary levels of GDF-15 and KIM-1 are depicted in the Additional file 1: Figure S1. Similar scatter dot plots representing in-hospital mortality are shown in Fig. 3 and in the Additional file 2: Figure S2.

The mean levels of the five different biomarkers, namely, NGAL, calprotectin, KIM-1, cystatin C, and GDF-15, in the serum and urine samples were compared, as shown in the Sepsis column of Table 3. Patients with sepsis had significantly higher levels of both serum and urinary NGAL (both P values < 0.001). Serum and urinary cystatin C (P value 0.016 and P value 0.046, respectively) levels were also significantly higher in septic patients. Other biochemical parameters showing statistical significance included serum albumin, creatinine, and CRP (P values 0.025, 0.004, and < 0.001, respectively). Septic patients also had statistically higher SOFA (P value 0.001) and APACHE II (P value 0.001) scores upon ICU admission.

As shown in the Septic AKI column of Table 3, the levels of serum and urinary NGAL (P values 0.001 and < 0.001, respectively) were significantly higher in patients with septic AKI than in those without. Although serum calprotectin (P value 0.049) showed statistical significance, urinary calprotectin (P value 0.102) was not significantly elevated in patients with septic AKI. The levels of serum albumin, creatinine, CRP, SOFA score, and the APACHE II score (P values 0.007, < 0.001, 0.008, < 0.001, and 0.001, respectively) all showed significant elevation in the septic AKI group when compared to the non-septic AKI group.

Patients with in-hospital mortality (n = 15) showed significantly higher levels of serum NGAL (P value 0.029) and urinary NGAL (P value 0.001) on ICU admission, as shown in the In-hospital Mortality column of Table 3. The SOFA and APACHE II scores, serum albumin, and creatinine also showed statistical significance (P values 0.003, 0.004, 0.003, and 0.027, respectively).

Table 4 summarizes the area under the ROC curves (AUROC) of significant variables regarding septic AKI and in-hospital mortality. In predicting septic AKI, serum and urinary NGAL showed an AUROC of 0.991 and 0.915, respectively. Serum calprotectin showed an AUROC of 0.889. Other parameters with statistical significance in predicting septic AKI included serum creatinine, CRP, SOFA, and APACHE II scores, with the exception of serum albumin.

In predicting in-hospital mortality, serum and urinary NGAL gave a statistical significant AUROC of 0.768 and 0.780, respectively. Similarly, the SOFA and APACHE II scores showed an AUROC of 0.774 and 0.762, respectively. Albumin also showed statistical significance in predicting in-hospital mortality, even though no statistical significance was observed in predicting septic AKI. Of great interest, a combination of serum NGAL, serum calprotectin, and SOFA score presented an AUROC of 1.000 (P value < 0.001) for septic AKI, while a combination of serum NGAL, urinary NGAL, and SOFA score gave an AUROC of 0.911 (P value < 0.001) for in-hospital mortality.

The ROC curves of individual plasma and urinary biomarkers along with the SOFA and APACHE II scores are shown in Fig. 4a, c, for the prediction of septic AKI and in-hospital mortality, and a combination of biomarkers and SOFA score is shown in Fig. 4b, d.

Further analysis was performed within the sepsis cohort, as shown in Additional file 3: Table S1. In predicting septic AKI, serum and urinary NGAL and serum calprotectin appeared to be statistically significant with an AUROC of 0.981, 0.885, and 0.962, respectively. A combination of serum NGAL, serum calprotectin, and SOFA score gave a high AUROC of 1.000 (P value 0.003). In predicting in-hospital mortality, similarly to what was demonstrated previously, a combination of serum NGAL, urinary NGAL, and SOFA score gave an AUROC of 0.963 (P value 0.001). The ROC curves of a combination of biomarkers and SOFA score in predicting septic AKI and in-hospital mortality are demonstrated in Additional file 4: Figure S3A and B.