Late Acute Kidney Injury is a Poor Prognosis Sign in Severe Burns: A Retrospective Cohort Study

Acute kidney injury (AKI) is a morbid complication and the main cause of multiple organ failure and death in severely burned patients. The objective of this study was to explore the epidemiological characteristics, the risk factors, and impact of both early and late AKIs, respectively. This retrospective study was performed with prospectively collected data of severely burned patients from the Institute of Burn Research in Southwest Hospital during 2011-2017. AKI was diagnosed according to Kidney Disease Improving Global Outcomes (KDIGO) criteria (2012), and it was divided into early and late AKIs depending on its onset time (within the rst 3 days or >3 days post burn). The baseline characteristics, clinical data, and outcomes of the three groups (early AKI, late AKI and non-AKI) were compared using logistic regression analysis. Mortality predictors of patients with AKI were assessed.

APACHE II, SOFA, and some laboratory values when AKI was diagnosed were also collected, such as: the serum creatinine (sCr), blood urea nitrogen (BUN), cystatin C, myohemoglobin, mean hourly urine output, and arterial blood gas analysis including pH, oxygenation index (PaO 2 /FiO 2 ) and . Occurrence time and duration of AKI were recorded as well.
De nition AKI diagnoses were made according to the KDIGO classi cation. KDIGO criteria for the diagnosis of AKI include [14]: an absolute increase in sCr of ≥0.3 mg/dL (≥26.5 μmol/l) within 48h, or an increase in sCr 1.5 times the baseline value (or rst measurement), or UOP <0.5mL/kg/hour for more than 6 hours. Baseline sCr was de ned as the the lowest value within hospitalization. The area of burn wound was assessed by "Chinese rule of nine", and the burn depth by the "Three degree and four categories" [12].
Sepsis and septic shock were de ned according to the "International Guidelines for Management of Sepsis and Septic Shock: 2016" [15].

Statistical analysis
The quantitative data were expressed as mean (SD) or median (interquartile ranges, IQR) according to their distributions, while categorical data were expressed as counts (proportions). Then we created a binary outcome variable for AKI: a value of 2 without AKI, 0 when there was early AKI (0-3 days post burn), or value of 1 when there was late AKI; also, binary outcome variables for mortality, and other clinical status. This relied on the fact that early AKI mostly occurs within the 3 days after severe burn, which renders the predictive model for AKI [1]. All statistical analyses were performed using SPSS (Version 25; IBM, Armonk, NY, USA).
The Student's t-test was applied if a normal distribution was detected, while the Mann-Whitney U test was applied for non-normal distributions. Wilcoxon rank sum test or Pearson χ 2 test for nonparametric variables was used when appropriate.
Risk factors in early AKI and non-AKI, late AKI and non-AKI groups were evaluated in univariate analysis rst. Then multivariate logistic regression was applied to the relevant variables from the clinical/physiological perspective. The odd ratios (OR) and 95% con dence intervals (CI) were estimated.
Moreover, ROC curves were built for myohemoglobin associated with AKI when the AUC-ROC was reported. Meanwhile, the sensitivity, speci city, positive and negative predictive values, and positive and negative likelihood ratios (PPV, NPV, PLR and NLR) were calculated for myohemoglobin. The Youden index (sensitivity + speci city -1) was applied to de ning the best threshold.
To assess the association between AKI and subsequent mortality, Kaplan-Meier estimate of survival was constructed to compare 90-day survival between each group via strati ed log-rank test. Mortality predictors of patients in early AKI and late AKI groups were assessed using multivariable Cox proportional hazards models. Hazard ratios (HR) with corresponding 95% con dence intervals (CI) were calculated. All testing was two-tailed, and statistical differences were considered signi cant if p < 0.05.

Sample size calculation
A previous study on building risk prediction models demonstrated that the calculation of a cohort sample size would precisely provide the required sample size [16]. However, the AKI prevalence in the institute database has not been estimated before. Given the reported prevalence of AKI cohort percentages of severely burned patients ranging from 1 to 36% [1], all patients available in the database over the study period (648 severely burned patients) were included to robustly estimates even when AKI prevalence was as low as 3%.

Results
General characteristics and incidence of AKI Totally, 648 patients were severely burned and enrolled, from whom 9 patients were excluded because of unspeci ed prehospital information and 2 patients because of CKD before burns. There were 20 patients discharged without a cure and missing to follow-up (early AKI 12, late AKI 1, and non-AKI 7). To avoid exclusion of a large proportion of patients, these 20 patients underwent analysis of risk factors but were excluded from the analysis of clinical outcomes. At last, 637 patients including 402 non-AKI patients and 235 AKI patients were included into the analysis. Of them, 16 patients developed both early and late AKI, and were recorded in both early and late groups. Thus, the AKI group contained 251 events (Fig. 1). The incidence of AKI was 36.9% (early AKI 29.4%, late AKI 10.0%). Characteristics of the patients are presented in Table 1. Most patients were middle-aged (45±13.2) and male (499, 76.1%), and 201 (30.1%) patients developed hypovolemic shock of early burn. The median value for TBSA, full-thickness burns of TBSA, and ABSI was 45.7% TBSA, 15.4% TBSA, and 9.9, respectively.
The baseline values of clinical parameters at onset of AKI are summarized in additional le 1. The average time of initial diagnosis of early AKI was 19 hours (the earliest time, 3 hours) after burn, while 19 days in late AKI group. Half of the patients with late AKI were unable to restore kidney function, accounting for a signi cant higher proportion than those with early AKI (18.19%, p<0.0001). For patients with renal function recovery, the mean AKI duration was 4.2 days in early AKI patients, which was signi cantly shorter than that in late AKI patients (6.5 days). Incidence of late AKI was lower than that of early AKI, but late AKI patients had more severity of illness and worse prognosis. On the day of AKI diagnosis, patients with late AKI had a higher illness severity, APACHEII score, SOFA score, bicarbonate radical, serum creatinine, BUN and cystatin C levels, than those with early AKI All patients with late AKI were admitted into ICU of burn, while 30 patients with early AKI were not. Moreover, the mean hourly urine output of patients with late AKI (100 ml/hour) did not decrease compared to normal adult urine output, while it decreased sharply in early AKI (40 ml/hour, p<0.001); pH, PaO 2 /FiO 2 and lactate in arterial blood had no signi cant differences between the two groups.
Furthermore, the prevalence showed that a correlation existed between AKI and age or TBSA (Table 2).
We noted that as the TBSA and age of the patient increased, the incidence of AKI did as well. Speci cally, 79% of the patients with burn >80% TBSA developed AKI, and over half of the patients aged above 60 years developed AKI (51.4%).

Multivariate analysis of risk factors
To eliminate the in uence of confounding factors, we analyzed the results by unordered multinomial logistic regression with the variables with P<0.1 from Table 1, and the regression results are presented in Table 3. The following factors were associated with AKI and considered as independent risk factors associated with early and late AKI: male, age, TBSA, full-thickness burns of TBSA, hypertension or/and diabetes, hypovolemic shock of early burn, and tracheotomy. Sepsis had the greatest in uence on late AKI occurrence (OR=12.831, p<0.001), and increased the risk of late AKI by 11.8 times, followed by hypovolemic shock of early burn (OR=10.232, p<0.001), which increased the risk of late AKI by 9.2 times. However, decompression escharotomy was a protective factor for early AKI (OR=0.452, p=0.004) and late AKI (OR=0.189, p<0.001), which reduced the risk by nearly 55% and 81%, respectively. The AUC-ROC of the model was 0.812 (0.738-0.831) for early AKI and 0.833 (0.818-0.853) for late AKI.

Myohemoglobin and prediction of AKI for severe burn patients
The discriminating power of myohemoglobin had an AUC-ROC of 0.79 (0.75-0.82) to predict all AKI (Fig.   2). Results of additional analyses, including sensitivity, speci city, PPV, NPV, PLR, NLR and optimal cutoff, are available in additional le 2.

Discussion
In this retrospective observational study, it was found that rstly, AKI was still a serious issue in severely burned patients with a 38.4% occurrence, and AKI was independently associated with a threefold risk of mortality in severely burned patients. Second, it was obvious that AKI occurred early, with 74% of AKI diagnosed within the rst 3 days after severe burn injuries, but late AKI had higher mortality. Third, the AKI predicting model was proved to include age, gender, TBSA, full-thickness burns of TBSA, chronic comorbidities (hypertension or/and diabetes), hypovolemic shock of early burns, inhalation injuries, and myohemoglobin. Tracheotomy and sepsis were independent risk factors for AKI.
AKI is a common complication in severely burned patients. The incidence of AKI in burned patients varied from 1% to 36% depending on the different population studied and diagnostic criteria, and the mortality among burned patients with AKI was 28% to 100% [17]. Other studies have shown that AKI occurred in 53.3% of patients with severe burn injuries, and patients with AKI had a mortality rate of 34.4% [4]. In this study, the incidence of AKI was 36.9% in the patients with burn ≥30% TBSA; the total mortality was 32.3% in the patients with AKI, highest in the patients with late AKI with the mortality up to 56.3%, which was higher in burn patients than in general population [5]. Therefore, it is crucial to early prevent and diagnose AKI in severely burned patients. However, in other study, we found the incidence and mortality of AKI have not decreased in the consecutive six years. Reducing the incidence of AKI and improving the prognosis of patients with AKI are still facing signi cant di culties and challenges.
In multinomial logistic regression, the study found that age, gender (male), TBSA, full thickness burns of TBSA, hypertension or/and diabetes, tracheotomy and hypovolemic shock of early burns were independently associated with both early and late AKI development, and sepsis was an independent risk factor for late AKI. However, BMI and drink, smoke, ABSI, etiology of burns and inhalation injuries had no in uences, which is similar to previous studies [18][19][20]. The risk of the aged and male patients suffering early and late AKI increased by over one to ve folds, which is in agreement with some studies [21][22][23], while others found gender is not a predisposing factor for AKI [10]. Considering that more male patients may be engaged in electrical works and suffered electrical injuries which may be a risk factor for AKI, this result was not changed after analyzing the cases of thermal burns (data not shown). The severity of burns is largely determined by burn area and depth. Major burn size is associated with multiple organ dysfunction and poor prognosis, and burn size is an independent predictor of acute renal failure [13]. In this study, TBSA and full-thickness burns of TBSA strongly correlated with AKI. Patients with hypertension or/and diabetes underwent chronic pathological changes in organs including the heart, brain, kidneys and blood vessels causing deterioration of renal function, and were more predisposed to AKI [17]. Inhalation injury was a severe complication of burn injuries, and was often associated with AKI [19]. Though univariate analysis in this study showed signi cant difference in inhalation injury between the groups, there was no evidence of an independent inhalation injury effect after adjusting for potential confounders. This result is consistent with recent study [24]. This could be because most mild to moderate inhalation injuries in this study did not develop to ARDS, which was proved to be a risk factor for AKI. ARDS is an important pathophysiologic link between acute lung injury and AKI and the existence of the "lung-kidney crosstalk" [24][25][26]. Tracheotomy was an independent risk factor for early AKI (OR=2.335, p=0.003) and late AKI (OR=3.708, p=0.003). This can be explained as that the patients with tracheotomy were more seriously ill with inadequate oxygenation and needed mechanical ventilation support. Meanwhile, there may be complications of secondary pulmonary infection induced by tracheotomy and mechanical ventilation. However, we also included a few patients to perform prophylactic tracheotomy for compression of the neck and laryngeal edema after burns.
There are different pathophysiological features and potential mechanisms for burn-related AKI at different stages after burns. Hypovolemic shock is the main complication of major deep burns in the resuscitative phase, with a hugely harmful in uence on the whole course of severe burns [27], when consequent sepsis and MODS increased the mortality accordingly. This study con rmed that hypovolemic shock of early burns is also a risk factor for AKI in severely burned patients. It increased the risk of early AKI, and also increased the risk of late AKI by 9.2 times. It has always been believed that renal hypoperfusion leads to ischemia and ischemia-reperfusion injury, and timely and aggressive uid resuscitation can reduce the incidence of AKI and improve prognosis [28]. We paid attention to the early uid resuscitation after burns, and drafted the TMMU protocol for uid resuscitation in the 1960s, which has been followed to save lives of innumerable burn patients, and until now it is still widely used in China [29]. However, aggressive uid resuscitation did not completely avoid AKI [30,31]. Furthermore, excess resuscitation was associated with some complications, including pneumonia, ARDS and elevated compartment syndromes [30,32]. This study found the earliest onset of AKI was 3h (or may be earlier) after burns, and AKI was still developed in burn patients who received rapid uid resuscitation, with normal or slight decrease of urine output. Hemorrhagic shock and septic shock in animal experiments demonstrated that immediate uid resuscitation was su cient to restore systemic blood pressure but failed to restore renal tissue oxygenation [33,34]. Therefore, etiology of early AKI is multifactorial, not solely focusing on the amount of uid received. Renal vasoconstriction by stress-related hormones (catecholamines, angiotensin , aldosterone, vasopression, etc.), in ammatory mediators, denatured proteins (myoglobin, myoglobinuria, free haemoglobinuria, etc.) are probably also associated with the occurrence of AKI [1]. It was also observed that sepsis was the main reason for late AKI 3 days after burns, and was an independent risk factor for late AKI, which increased the risks of late AKI by 11.8 times. This was consistent with a former study [36]. Interestingly, late AKI occurred before the change of urine volume. On the day of initial diagnosis, urine of patients with late AKI did not decrease (100ml/hour), but sCr increased (156 umol/L). It is likely that renal blood ow did not decrease, but the creatinine clearance has already been markedly reduced. This con rms that the in ammatory mediators and microcirculation dysfunction mainly caused by sepsis and infection contribute to the development of late AKI.
Myohemoglobin (myoglobin) is a harmful product of rhabdomyolysis, which is a clinical syndrome secondary to skeletal muscle injury [36]. Myohemoglobin is the main cause of renal failure. In severely burned patients, myohemoglobin was proved to have a very stable predictive value for AKI with pronounced sensitivity and speci city [37], as is also reported in other traumatic injuries [38].
It is interesting to note that decompression escharotomy of circumferential burns probably reduced the risk of AKI development. Circumferential eschar, generally resulted from third-degree burns, can cause compression of the underlying soft tissues as burn edema develops and lead to compartment syndrome [39]. With the increase of the compartment pressure, ischemia of the underlying tissues and the distal tissues will result in tissue and muscle necrosis, and lactic acidosis without timely decompression escharotomy. Toxins released from necrotic tissue are an important etiology of AKI. Decompression escharotomy should be considered earlier, especially in patients with very deep thermal and/or electrical injuries, as it may be effective in the prevention of AKI.
Several studies have emphasized that AKI is independently related to adverse clinical outcomes in trauma patients [5] and ICU patients [40]. In this cohort, AKI was independently associated with high mortality. Thus, renal dysfunction serves as an additional predictor of huge risk of mortality in severely burned patients.
This study is not a multicenter study, though the total number of AKI patients in the sample far exceeded the recommended number for building a model to provide predictions [16]. This study demonstrated an association between risk factors and AKI occurrence, but has not established causality. Fortunately, the single center study performs a better stability in burn patients' data managing, diagnosis criteria and treatment, which ensure that the data are more convincible than a multicenter database where subjective factors might be ampli ed.
A sensitive and speci c biomarker for an accurate early diagnosis of AKI is urgently needed even in the absence of subsequent dysfunction. The novel biomarker is probably a produced material of damaged kidney itself, just like the speci c cardiac troponin protein produced immediately when myocardial infarction occurs.

Conclusions
AKI which is independently associated with high mortality in severely burned patients, is still highly prevalent. Age, gender, TBSA, full-thickness burns of TBSA, hypertension or/and diabetes, tracheotomy, hypovolemic shock of early burns and sepsis are independent risk factors for AKI. Decompression escharotomy is clearly associated with decreased risk for AKI, especially for late AKI. It is crucial to adopt early effective and individualized prevention strategies for severely burned patients to prevent AKI from taking their lives.