With this retrospective observational study, we aimed to describe epidemiological and clinical characteristics and to identify factors associated with mortality in our group of critically ill pediatric patients who had received CRRT. The current study is one of the largest single-center studies in the literature, including 121 pediatric patients from our PICU. Our study provides a representative cross-section of a large and heterogeneous PICU population receiving CRRT including all diagnostic subgroups.
The use of CRRT has become an integral part of modern critical care and is used for a variety of clinical situations in critical ill children with renal and non-renal indications [1–5]. However, despite its increased use, identifying the optimal indication and the optimal timing remains a problem in clinical practice. Mortality rates for critically ill children receiving CRRT range from 27–57.2%, with some improvements over the last decade that may be associated with various factors, including increased use in non-critical children, better recognition of AKI in critically ill children, or increased safety due to technological refinements [3, 6–15]. The 29.8% mortality rate in our study stands as one of the lowest when compared to previous publications. However, it is apparent that children requiring CRRT represent a heterogeneous group, and previous studies have shown that outcome is closely related to underlying disease [3, 7–9, 13]. In our study, mortality in children with hemato-oncological disease or pneumonia/respiratory failure was very high (73.3% and 72.7%, respectively); whereas, children with primary renal disease had very low mortality rates (5.4%). These findings are comparable with data from the pediatric CRRT registry as published by Symons et al., Beschee et al., and Cortina et al. [3, 4, 13]. In our study, survival rate was higher among patients in whom CRRT was required for AKI without any other organ failure (94.1%). Hames et al. also found superior survival in patients treated for isolated AKI (85.7%) .
We found that age, sex and weight had no influence on mortality in the current study, similar to previous reports [3, 4, 20]. Non-survivors were found to have lower weight at PICU admission in a study by Hames and colleagues . The PRISM III score, which is calculated in the first 24 hours of PICU admission, is a frequently utilized method of prognostic assessment in critically ill pediatric patients. As expected, PRISM III scores were higher among non-survivors. However, studies evaluating recipients of CRRT in the PICU report varying results; while some studies have found relationships between PRISM III score and mortality, others have not [4, 8, 13, 19–21]. In two studies by Golden et al.  and Şık et al. , it was found that mortality was associated with PRISM III score that was re-calculated immediately prior to beginning CRRT treatment. Multiple studies have identified that the need for vasoactive support [10, 14, 15], need for mechanical ventilation [15, 19], presence of comorbid factors [3, 22], and MODS [10, 13] were risk factors of mortality in recipients of CRRT. In the current study, although we found higher PRISM III and vasoactive support score, and higher frequency of mechanical ventilation, comorbidity and MODS in the non-survivor group, we did not find independent relationships between mortality and any of these factors.
In the present study, the most common primary indication for initiating CRRT was electrolyte or acid-base imbalance, followed by AKI and %FO. Although the distribution of primary CRRT indications vary from study to study, they are generally in agreement with the aforementioned findings [13, 19, 21, 22]. We found no relationships between CRRT indications and mortality. Similarly, Şık et al.  also reported no relationship between mortality and CRRT indication. However, Cortina and colleagues  have reported that CRRT performed due to FO was associated with higher mortality rate compared to other indications.
In the literature, the median %FO values at the initiation of CRRT range from 6.3–21.0% [10, 13, 19, 21, 22]; in the current study we report a value of 6.7%. Multiple studies have demonstrated that higher FO at the time of CRRT is an independent risk factor for increased mortality and morbidity [6–9, 12]; however, conflicting results also exist [4, 19, 23]. In the present study, non-survivors were found to have a significantly higher degree of FO compared to survivors (3.0% vs 14.2%). Although this was not associated with mortality on multivariate analysis, we believe there should be continued emphasis on managing the fluid status of critically ill children. This may be especially true in patients with respiratory disease. When these characteristics were evaluated with regard to diagnoses, we found that non-survivors admitted with pneumonia/respiratory failure had a median fluid accumulation of 17.9%, which was a higher value compared the 7.0% in survivors. Thus, it is possible that this characteristic may have contributed to the high mortality rate for this subgroup of patients. In the current study, %FO was greater than 20% in 18.8% of survivors and 38.9% of non-survivors; however, statistical significance was not found in terms of mortality. It must be noted that several studies have reported > 20% FO as an independent risk factor for mortality; whereas others have not identified a relationship [13, 19, 22].
AKI is common in pediatric patients admitted to the PICU, and its frequency rises with increasing severity of illness [13, 24]. The recently published prospective multicenter study, Assessment of Worldwide Acute Kidney Injury, Renal Angina, and Epidemiology, found that AKI developed in 26.9% and severe AKI was present in 11.6% of critically ill children and young adults. The study concluded that severe AKI was associated with higher mortality rate . In our study, even though we found higher KDIGO stages (stage 2 + 3 AKI) in non-survivors compared to survivors (61.1% vs. 45.9%), statistical results showed no significance. Our results were similar to those reported by Cortina and colleagues who did not find any relationship between severe AKI and mortality .
The time to initiate CRRT is still controversial in the literature . The varying definitions of early versus late initiation further compounds this problem. Two randomized controlled trials in critically ill adults with AKI showed conflicting results. The “Effect of Early vs Delayed Initiation of Continuous Renal Replacement Therapy on Mortality in Critically Ill Patients with Acute Kidney Injury (ELAIN)” trial demonstrated lower mortality when patients received early CRRT (within 8 hours of diagnosis of KDIGO stage 2 AKI) compared to those who had been treated with a delayed strategy (within 12 hours of stage 3 AKI or no initiation) . However; the “Artificial Kidney Initiation in Kidney Injury (AKIKI)” trial did not show a difference in mortality between patients assigned to receive early versus late CRRT . For this trial, all patients had stage 3 AKI, and early initiators received RRT immediately after developing stage 3 AKI; whereas late initiation was performed only if they had an acute indication for CRRT. The detailed comparison of the two studies further supports our belief that the variation in definitions (or rather, a lack of definitive characterization) remains as an important problem. In our study, the median time of initiation was 7 hours, which, compared with other pediatric studies, is very early [13, 19, 21, 22]. This characteristic difference was due to the traditional approach taken our unit, which focuses on starting CRRT early in select patient groups, including those with sepsis and other acute clinical conditions (such as hemolytic uremic syndrome). In our study, univariate analysis revealed that non-survivors were started on CRRT later than survivors (with respect to time from PICU admission). However, multivariate analysis did not identify CRRT initiation time as a factor that contributed to mortality. The definition of early or late CRRT treatment varies in pediatric studies, and the majority of research on this topic has assessed the time from PICU admission to CRRT initiation in order to determine whether it is associated with mortality. However, results are conflicting. Some studies in pediatric patients have reported increased survival with early CRRT [12, 13], while others have found no relationship [4, 19, 21, 22].
Timing the initiation of CRRT from PICU admission can be somewhat arbitrary, as it does not appropriately categorize patients who arrive to the PICU with normal renal function and subsequently develop AKI later during the course of their PICU stay. Kaddourah et al.  performed a prospective, multicenter observational study evaluating RRT recipients among critically ill patients with AKI older than 12 years old. Late RRT, defined from PICU admission, was associated with increased mortality; but when it was defined based on a blood urea nitrogen threshold, there was no difference in mortality. Conversely, mortality was lower for late RRT initiation when timing was based on a serum creatinine threshold. Although very interesting and informative, this study did not evaluate stage of AKI as a factor. In a study by Hames and colleagues, the time until CRRT initiation from the development of stage 3 AKI was reported to be significantly longer in non-survivors. However, the time between peak creatinine level measurement and CRRT was reported to be unassociated with survival .
We also evaluated whether CRRT initiation time was influential on survival with regard to admission diagnoses. Patients with sepsis who had survived were found to have shorter time until CRRT initiation compared to non-survivors. This particular relationship was also identified in the study by Cortina and colleagues; however, their results did not show statistical significance . It has been previously reported that early CRRT may be utilized in patients with sepsis –even in the absence of AKI– in order clear inflammatory mediators, correct acid-base imbalance and prevent FO [13, 29].
In the present study, overall median CRRT duration was 3 days. Although median CRRT duration was longer in non-survivors, multivariate regression did not yield a significant difference between groups. This was consistent with the report of the “Prospective Pediatric Continuous Renal Replacement Therapy (ppCRRT) Registry” and another smaller study of 190 patients, both of which found no significant difference in terms of survival rate based on duration of CRRT [3, 12]. Furthermore, we also report a significantly longer PICU stay in non-survivors (7.0 vs 4.5 days), but multivariate regression again showed no significant effect on survival, similar to other pediatric series [20–22].
The relationships between survival and some pre-CRRT biochemical/hematological characteristics were also assessed. Platelet count was significantly lower in survivors compared to non-survivors. This was possibly due to the fact that the majority of patients with the lowest (5.4%) mortality rate (those with renal disease) had a diagnosis of hemolytic uremic syndrome –a disease that is characterized by low platelet count. Serum creatinine was also higher among those that survived; however, this can also be associated with the increase in creatinine levels in patients that were admitted with renal disease (who had very high survival). Other studies in the literature also report no relationship with creatinine and mortality in CRRT recipients [4, 19]. We only identified lactate as an independent risk factor in this study. Even though the literature reports different results, Cortina et al.  and Fernandez et al.  have not found an association between mortality and lactate levels. Furthermore, although lactate levels were higher among non-survivors in a study by Choi and colleagues , it was not found to be an independent risk factor on mortality. However, in an adult study comprised of CRRT recipients due to AKI development after cardiac surgery, it was reported that > 5 mmol/L lactate level at CRRT initiation was associated with mortality (94.1%) .
There are several limitations to this study. This is a retrospective analysis and is therefore reliant on accurate documentation in medical records. The fact that this was a single center study also limits generalizability, especially due to likely variations in practice and patient diagnoses, as standardized management guidelines do not exist for CRRT. Although the study population was relatively large, it is also heterogeneous and involves smaller subgroups, which limits the precision of our estimates and makes it difficult to draw general conclusions from the associations observed. Another limitation is the method of calculating of %FO, which was performed by evaluation of fluid from PICU admission to CRRT initiation, and it is apparent that some patients might have received larger fluid volume before admission to our unit. Insensible losses were also not accounted for in the fluid balance calculation. Another potential limitation is lack of data regarding the rate of fluid removal after initiating CRRT. This can be a potential factor that affects survival, ventilation time and length of PICU stay.