Meta-Analysis of Renal Replacement Therapy for Burn Patients: Incidence Rate, Mortality, and Renal Outcome

Background: Renal replacement therapy (RRT) was often needed by some severe burn patients with acute kidney injury (AKI). The primary aim of this study was to review incidence rate and mortality of RRT in severe burn patients. Second aims were to review RRT complications and renal outcome. Methods: We searched multiple databases for studies published between 1 January 1960 and 31 December 2019. Studies about adult populations with burn injury, providing epidemiologic data on prevalence or mortality of RRT, were included. Results: A total of selected 57 studies, including 27,437 patients were enrolled in our analysis. The prevalence rates of RRT were 8.34% (95% CI 7.18–9.5%) in all burn patients and 37.05% (95% CI 29.85–44.24%) in AKI patients. The mortality of all burn patients with RRT was 65.52% (95% CI 58.41–72.64%). The prevalence rates of RRT in sample size≥100 group were 6.86% (95% CI 5.70–8.03%), which was lower than that of <100 group (17.61%, 95% CI 13.39–21.82%). With the increase of TBSA, the prevalence of RRT may have the increasing trend. The prevalence rates of RRT in Asian group was 12.75% (95% CI 9.50–16.00%), which was higher than that of European (10.45%, 95% CI 7.30–13.61%) and North America group (5.61%, 95% CI 4.27–6.95%). The prevalence rates of RRT in 2010–2019 group was 12.22% (95% CI 10.09–14.35%), which was higher than that of 2009–2000 group (5.17%, 95% CI 2.88–7.46%). The prevalence rates of RRT in 1989 and before group was the lowest, which was 1.56% (95% CI 0–3.68%). However, there was no significant correlation between the year of publication and the mortality of burn patients with RRT. Dialysis-requiring AKI in burn patients could increases the risk of chronic kidney disease progression and end-stage renal disease. About 35% of RRT patients need to maintain haemodialysis temporarily, even if they survive and leave hospital. Conclusions: The prevalence rate of RRT is about 6–8%; approximately, one-third of burn patients with AKI need RRT. The prevalence rate of RRT increased over time, but the mortality did not change. The prevalence rates of RRT in Asian group was higher than that of European and North America group.


INTRODUCTION
Acute kidney injury (AKI) is one of the common complications of burn patients that seriously threatens the life of patients and increases the length of stay, intensive care unit (ICU) length of stay and treatment costs (1,2). In 2010, metaanalysis reported that the prevalence of AKI in burn patients ranged from 16 to 26.6% (3), depending on the severity of the burns and on the definition of AKI. The prevalence of AKI in burn patients ranged from 18.4 to 47.4% with the RIFLE standard (2). Burn patients with AKI had a significantly increased risk of death. The mortality of burn patients with AKI was 16.95-100%, significantly higher than that of their control group .41%) (4)(5)(6)(7).
However, the prevalence and mortality of renal replacement therapy (RRT) for burn patients are still unclear. A meta-analysis in 2010 conducted a subgroup analysis of the prevalence and mortality of RRT in burn AKI patients. The results showed that the prevalence of RRT was 3.2% in all burn patients and 27.1% in burn patients with AKI. The mortality of burn patients with RRT was as high as 80% [95% confidence interval (CI) [72][73][74][75][76][77][78][79][80][81][82][83][84][85][86][87][88].6%] (3). Moreover, the updated meta-analysis of burn patients admitted to the ICU showed that the prevalence of RRT in all burn patients was 12%, and the mortality was 74% (95% CI 58-87%) (8). A recently published multicenter observational study showed that the mortality rate of RRT in severe burn patients has been reduced to about 50% (9). The mortality of burn patients with RRT was still very high. However, due to the different directions of attention, the literature of RRT for burn patients included in these studies is incomplete. At present, there is a lack of meta-analysis of RRT for burn patients. The primary aim of this study was to review incidence rate and mortality of RRT in burn patients. Second aims were to review RRT complications and renal outcome.

METHODS
This systematic review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyzes and the Meta-Analysis of Observational Studies in Epidemiology guidelines (10,11).

Eligibility Criteria
Studies on adult populations with burn injury, providing epidemiologic data on prevalence and mortality of RRT, were included. The adult (>16 years of age) patients with burns covering 10% or more of the total body surface area (TBSA) were enrolled. Moreover, TBSA used in the subgroup analysis. Studies conducted only in patients with chemical or electrical burns were not included because of the different pathophysiologies. We excluded studies with a sample size of <10. Other blood purification techniques, such as plasma exchange (non-RRT), are not included in this study.

Search Strategy
We used the following terms for standard medical subject headings and free-text words: burn, burns, renal, kidney, kidney injury, kidney diseases, renal insufficiency, renal failure, and kidney failure. We also reviewed the references cited in all the studies selected for review. The search was performed in January 2020 and considered the period 1960-2019. The year of publication used in the subgroup analysis. An extensive search of literature published was conducted using the databases PubMed/MEDLINE, Embase, Science Citation Index (Web of Science), and the Cochrane Central Register of Controlled Trials (CENTRAL) database. Two searchers (D.ZY. and C.FK.) conducted the search independently. If one of the searchers thought that the article was appropriate, the study was selected.

Study Selection
Two reviews (D.ZY. and C.FK.) independently screened studies for eligibility according to study selection criteria. The inclusion criteria were randomized controlled trial (RCT), case-control and cohort studies. Cross-sectional studies were excluded because they cannot clearly identify the causal relationship between observed indicators and diseases. Only articles in English, Japanese or Chinese were included. Populations with pediatric patients, animal studies, case reports and reviews were excluded. Any disagreement between reviewers were resolved by consensus. RRTs included slow low-efficiency dialysis, intermittent haemodialysis (IHD), peritoneal dialysis (PD), and continuous renal replacement therapy (CRRT). We excluded the study of plasma exchange in burn patients because the mechanism and mode of treatment of plasma exchange are different from those of dialysis. We excluded the lack of mortality studies aimed at studying antibiotic pharmacokinetics in burn patients with RRT. We excluded the pre-existing comorbidity for RRT such as end-stage renal disease. For studies that gave multiple mortalities, we took the longest mortality, such as 28, 60, 90, or in-hospital mortality.

Study Quality
We assessed risk of bias for RCTs using the Cochrane Collaboration's tool for assessing bias (12). We assessed risk of bias for case-control and cohort studies using the Newcastle-Ottawa Scale (NOS) (13). The NOS allocates nine points for quality of the selection (four items, four points), comparability (one item, two points), and outcome or exposure (three items, three points). Publication bias was assessed by creating and examining funnel plots. The robustness of the results was evaluated using sensitivity analyzes. Each study included in this review was assessed for quality as good (7)(8)(9), moderate (4-6), or poor (≤3) based on scores.

Data Extraction and Statistical Analysis
Two independent reviewers (D.ZY. and C.FK.) extracted the following information from the included studies: title, author, year, journal, study design, nationality, sample size, definition of AKI, number of patients with or without AKI, number of patients with or without RRT, RRT details (modalities, device and manufacturer, dose, anticoagulants), and results (mortality of AKI patients, mortality of RRT patients). We also reported other outcomes (RRT complications, renal outcomelong-term dialysis, temporarily required dialysis, chronic kidney disease progress). All ambiguities in data extraction were double-checked and resolved. Relative risks and 95% confidence intervals were obtained using a random effects model. I 2 derived from the chi-squared test was used to evaluate the heterogeneity across the included studies. An I 2 <50% indicated that there was no significant heterogeneity (14). Sensitivity analysis was performed by sequentially removing each individual study. We assessed publication bias by constructing a funnel plot. Pearson or Spearman's correlation was used to analyze correlations. A twotailed P-value of <0.05 was considered statistically significant. All statistical analyzes were performed using Review Manager version 5.3, SPSS 24.0 and R 3.5.1 software were used for metaanalyzes of prevalence and mortality.

Quality Assessment and Study Characteristics
We screened and evaluated 4,720 studies, assessing 194 for eligibility. The selected 57 studies, including 27,437 patients, were enrolled in our analysis (Figure 1) (4-7, 9, 15-65). There were 4,345 burn patients with AKI and 1,754 burn patients with RRT. A total of 93.5% of the burn patients with RRT were AKI patients. Figure 1 shows a flow chart of the identification and selection of the studies.

Prevalence and Mortality of RRT in Burn Patients
We analyzed 45 literatures that reported the prevalence of RRT in burn patients ( In order to balance the bias of small sample study (n < 100), we did subgroup analysis for the study with sample size >100 (Supplementary Table 2). The prevalence rates of RRT in these studies were 6.86% (95% CI 5.70-8.03%), which was lower than that of sample size<100 group (17.61%, 95% CI 13.39-21.82%). The prevalence rates of RRT in sample size ≥1,000 group were 2.52% (95% CI 1.02-4.02%). However, one third of the studies in sample size ≥1,000 group were from the 1980s, and the prevalence rates of RRT may be reduced because RRT was still in its infancy at that time.
The TBSA was used for subgroup analysis (≥10%, ≥20%, ≥30% or second and third degree burns >10%, ≥40% or second and third degree burns >20%). The results showed that, with the increase of TBSA, the prevalence of RRT may have the increasing trend (Supplementary Table 3). The prevalence rates of RRT in  Table 5).
According to the three diagnostic criteria of RIFLE, AKIN, and KDIGO, the prevalence of RRT was KDIGO < RIFLE < AKIN, and that of mortality was KDIGO < AKIN < RIFLE. The prevalence of RRT was 15.97% (95% CI 9.51-22.44%) and that of mortality was 55.29% (95% CI 39.46-71.12%) in the six literatures with KDIGO classification as the diagnostic standard, which was lower than other AKI diagnostic standards.
There was no significant correlation (r = −0.224, P = 0.159) between the year of publication and the mortality of burn patients with RRT (Supplementary Figure 2). According to the year of publication, the patients were divided into four subgroups (Supplementary Figure 3) from 2010 to 2020, 2000 to 2009, 1990 to 1999, 1989 and before. The mortality of the 2010-2020 group was 60.42 ± 25.35%, that of the 2000-2009 group was 61.55 ± 23.29%, that of the 1990-1999 group was 87.33 ± 8.44%, and that of the 1989 and before group was 63.52 ± 25.05%. There was no significant difference between groups (P = 0.139). After 2010, three studies still reported that the mortality of RRT patients was more than 90% (35,58,65). The prevalence rates of RRT in 2010-2019 group was 12.22% (95% CI 10.09-14.35%), which was higher than that of 2009-2000 group (5.17%, 95% CI 2.88-7.46%). The prevalence rates of RRT in 1989 and before group was the lowest, which was 1.56% (95% CI 0-3.68%). The results showed that the prevalence rate of RRT increased over time (Supplementary Table 6).

DISCUSSION
Acute kidney injury is one of the common complications in burn patients. The incidence of AKI varied from 16% to 26.6% according to the definition of AKI in different populations (3). The incidence of AKI in burn ICUs was 38% (30-46%) (8). RRT was often needed by some severe burn patients with AKI. According to the subgroup analysis in the previous meta-analysis of burn patients with AKI, the proportion of all burn patients requiring RRT was ∼3% (3). The proportion of burn patients admitted to the ICU was 12% (8-16%) (8), and the mortality was 74% (95% CI 58-87%) (3). A recently published multicenter observational study showed that the mortality rate of RRT in severe burn patients has been reduced to about 50% (9). The mortality of burn patients with RRT was still very high. Such high mortality rates were the reasons we wanted to conduct a metaanalysis of RRT in burn patients. A total of 27,437 burn patients were enrolled in this study from 57 literatures from 1979 to 2019. The results showed that the prevalence rates of RRT were 8.34% (95% CI 7.18-9.5%) in all burn patients and 37.05% (95% CI 29.85-44.24%) in AKI patients. These data are higher than those reported in the subgroup of previous studies (3,8). Due to the different purposes of previous studies, the number of RRT studies included in previous studies was significantly lower than that of this study.
The study of small sample size may enlarge the observation effect, which was first described by Sterne et al. (71). Small trials are more likely to report larger beneficial effects than large trials in critical care medicine (72). This study also found that compared with the large sample size study (n ≥ 100), the small sample size study may significantly enlarge the observation effect, that is, the prevalence rates of RRT. However, since the prevalence rates of RRT is not high in all burn patients, and some small sample size studies can provide more information on mortality, adverse reactions and renal outcome in RRT patients, we did not exclude small sample study, but instead, we used subgroup analysis to list them separately.
With the increase of TBSA, the risk of AKI is greater, and the risk of RRT may also be greater. High TBSA was a risk factor for AKI in burn patients (8). Our results showed that, with the increase of TBSA, the prevalence rates of RRT may have the increasing trend. The prevalence rates of RRT in TBSA ≥40% group was the highest. In a retrospective study of the aluminum dust explosion accident in Kunshan factory, 157 patients with severe burns were included, with an average TBSA more than 90%. In this study, the incidence of AKI was 56.7%, while the prevalence rates of RRT was 52.2% (67). In the study location subgroup analysis, the prevalence rates of RRT in Asian group was 12.75%, which was higher than that of European and North America group. The reason for the increased the prevalence rates of RRT in Asian group may be closely related to the large number of small samples studies (about 50%, 6/13) and the large TBSA of patients (TBSA ≥40% goup, 2/5).
With the improvement and popularization of RRT, more and more severe burn patients with AKI can be treated with RRT. Our results showed that with the progress of the times, the prevalence rates of RRT in burn patients is gradually increasing from <2% before 1980s to 12.22% now. Before CRRT was popularized, some severe burn patients with AKI were unable to tolerate IHD because of hemodynamic compromise. Chung et al. reported in 2008 that their center started CRRT in November 2005. Compared with the CRRT group, none of the 16 patients in the consecutive historical control group matched with TBSA and injury severity score were placed on haemodialysis (44). In Chinese mainland, only 15 of 6,050 burn patients underwent RRT in the 1980s, and most of them were PD. About 1/4 of the patients receiving PD developed abdominal infection (25). In 2012, the number rose to 6.31%, and all of them received CRRT treatment (21). Moreover, economic reasons may also affect whether burn patients start RRT. A survey of nephrologists and critical care physicians in 189 different hospitals in China showed that one of the reasons prohibiting patients from getting RRT were high therapy costs (73). With the development of economy and the improvement of medical insurance reimbursement policy in recent years, the prevalence rate of RRT in severe burn patients with AKI may also increase. CRRT could make more severe burn patients with an indication for RRT and hemodynamic instability have the opportunity received RRT treatment. However, RCT and meta-analysis studies in critically ill patients did not find that CRRT had a survival benefit advantage compared with IHD and PD in hemodynamic stable patients (74,75). There was no significant correlation between the year of publication and the mortality of burn patients with RRT in this study. Patients in the 1980 and 1990 groups received peritoneal dialysis or haemodialysis. Compared with the 1989 and before group and 1990-1999 group (received IHD or PD), most patients in the 2010-2020 group received CRRT, but there was no significant difference in mortality between groups.
RIFLE (76), AKIN (77) and KDIGO (78) are three commonly used AKI grading standards after 2004. The KDIGO diagnostic standard proposed in 2012 had been reported to improve the early diagnosis of kidney injury (59,79) and to reduce the missed diagnosis rate (80), which was conducive to the early diagnosis and treatment of AKI. Our results showed that, compared with RIFLE and AKIN, the prevalence and mortality of RRT in KDIGO were relatively low. We speculate that early diagnosis and early treatment may help some burn patients with AKI avoid further deterioration of renal function, reduce the need for dialysis, and improve the prognosis of patients. Ren et al., defined AKI according to KDIGO guidelines by 48 h after admission. That prospective study reported 11.2% (11/95) AKI and only 35.4% mortality in the AKI group (64).
AKI in burn patients not only increases mortality (81), but also increases the risk of CKD progression and end-stage renal disease (69,70). When severe burn patients have dialysis-requiring AKI, ∼35% of the patients need to maintain haemodialysis temporarily, even if they survive and leave the hospital. Among them, 10% of the surviving patients need haemodialysis for more than half a year. Therefore, for burn AKI or dialysis-requiring AKI survivors, it is recommended to monitor the renal function regularly within 1 year after discharge and to avoid the use of nephrotoxic drugs, excessive diuresis, and diarrhea (82), to reduce the progress of CKD and to avoid the recurrence of AKI.
This systematic review has several limitations. First, there was significant heterogeneity in most results of RRT prevalence and mortality in this study. The clinical heterogeneity of this study may be due to the differences in the study population, RRT modalities and the end-point indicators observed in the study. The inclusion and exclusion criteria were different among different studies. There were controversies on modalities, timing, and dosage of RRT, and the standards reported in different literatures are not the same. Outcome variables, such as mortality, also varied from study to study. Although we conducted a series of subgroup analyzes on AKI diagnostic criteria, ICU, different mortality criteria, TBSA, sample size, study location, and publication year, the results still showed heterogeneity. Because we were unable to obtain the original data for each patient, we cannot completely control the above confounding factors. Therefore, we used the random effect model to analyze the results. Second, we only included documents in English, Japanese and Chinese, and there may be omissions in documents reported in other languages, such as Russian. There was no literature from Russia in the studies we included. However, 57 articles and nearly 30,000 burn patients from five continents were included in this meta-analysis, most of which were in North America, Asia, and Europe. The inclusion of a large number of literatures and patients can ensure that the results of this study were well-representative, especially in North America, Asia, and Europe. Third, this study did not analyze the starting time of RRT for burn patients with AKI. At present, the timing of RRT for critically ill patients is still controversial. At the beginning of this study, we plan to analyze the timing of RRT in burn patients. However, even if 57 articles were included, there is still a lack of head-to-head study on the starting time of RRT for burn patients with AKI. There was only one cohort study that was divided into groups according to the AKI level of patients at the beginning of CRRT. The early CRRT group was defined according to the patients who started CRRT at the risk stage of the RIFLE criteria. The results showed there were no significant differences in mortality by the severity of AKI at the time of CRRT initiation (37). We suggest that more head-to-head cohort studies or RCT studies can be designed in the future regarding the timing of RRT in burn patients. Finally, the current study is not registered, and there may be a small deviation, but we still strictly follow the steps of systematic review.
Our meta-analysis had a number of strengths. In this study, 57 articles from 1979 to 2019 were included, involving nearly 30,000 burn patients. It has obvious representative advantages for the prevalence and mortality of RRT for burn patients with high statistical variability. Due to the large number of included literatures, we were able to conduct a series of subgroup analyzes on AKI diagnostic criteria, ICU, different mortality criteria, TBSA, sample size, study location, and publication year. As a result of a series of subgroup analysis, we found that The prevalence rate of RRT increased with TBSA and time. The prevalence rates of RRT in Asian group was higher than that of European and North America group.

CONCLUSIONS
The prevalence rate of RRT is about 6-8%; approximately, onethird of burn patients with AKI need RRT. The prevalence rate of RRT increased over time, but the mortality did not change. The prevalence rates of RRT in Asian group was higher than that of European and North America group.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
GC and JL contributed to the conception and design of the study. ZD and FC selected the studies, extracted the data, performed the statistical analyzes, and data presentation. ZD evaluated the study quality. The manuscript was drafted by ZD, with assistance by FC and JL, and the manuscript was critically revised by XC and GC. All authors contributed to the article and approved the submitted version.

FUNDING
This study was supported by National Key R&D Program of China (2018YFA0108803), Logistics research project (BLB19J009), Science and Technology Project of Beijing, China (D181100000118004, Z161100000516225, and D1771100002817002), and Natural Science Foundation of China (NSFC) (81600548 and 81670694). They have no role in study design, data collection, data analysis, data interpretation, or writing of the report.