Renal replacement therapy is independently associated with a lower risk of death in patients with severe acute kidney injury treated with targeted temperature management after out-of-hospital cardiac arrest

This was a retrospective analysis of the prospectively collected multicentre observational cohort study. Data were collected from the Korean Hypothermia Network (KORHN) prospective registry (KORHN-PRO) between October 2015 and December 2018. The KORHN is a multicentre clinical research consortium for TTM in South Korea. In total, 22 academic hospitals participated in the KORHN-PRO.

All adult patients (≥ 18 years) with OHCA, regardless of the cause of arrest, who were unconscious (Glasgow Coma Scale score < 8) after the return of spontaneous circulation (ROSC) and treated with TTM, were enrolled in the registry. Patients with active intracranial bleeding, acute stroke, known limitations in therapy and do-not-attempt resuscitation order, known pre-arrest cerebral performance category (CPC) 3 or 4, known disease making 6-month survival unlikely, and body temperature < 30 °C on admission were excluded.

Enrolled patients received care for post-cardiac arrest syndrome (PCAS) according to standard operating procedures regarding OHCA at each hospital. The principal investigator of each participating hospital collected data from the hospital records of OHCA survivors treated with TTM. Using a telephone survey, the independent data input committee investigated the survival and neurological status 6 months after ROSC. If patients died before 6 months, the death date was recorded.

The primary outcome was 6-month mortality and the secondary outcome was CPC at 6 months. We examined the following variables: age, sex, body weight, medical history (hypertension, heart failure, diabetes mellitus), arrest location (home, workplace, street, public building, nursing home, unknown), witnessed arrest, bystander cardiopulmonary resuscitation (CPR), time from collapse to CPR, time from CPR to ROSC, epinephrine dose, initial rhythm assessed by emergency medical service personnel in the field (ventricular fibrillation, pulseless ventricular tachycardia, pulseless electrical activity, asystole, unknown shockable, unknown non-shockable, unknown), arrest cause (medical or unknown, trauma, submersion, drug overdose, asphyxia, hanging), post-ROSC lactate, post-ROSC shock (cardiovascular sequential organ failure assessment score ≥ 1 on day 1), targeted temperature of TTM (33 or 36 °C), duration of TTM (24 or 48 h), adverse events during the 7 days following ROSC (seizure, bleeding, sepsis, hypoglycaemia [blood glucose< 60 mg/dl], sustained hyperglycaemia [blood glucose > 180 mg/dl for > 4 h], tachycardia [> 130/min], and bradycardia [< 40/min]), RRT modality (continuous renal replacement therapy [CRRT] or intermittent haemodialysis), RRT initiation time, reasons for initiating/terminating RRT, dialysis dependence at discharge, and serum creatinine (SCr). The KORHN-PRO required input of data on SCr daily until 7 days after ROSC. If the SCr was checked twice or more in 1 day, the highest value was collected in the registry.

We defined AKI based on the diagnostic criteria stipulated in the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines [12]. Because we did not have data on patients’ prior kidney function, we used the “modification of diet in renal disease study” formula for glomerular filtration rate (GFR) to estimate baseline SCr using the lower end of the normal range of GFR [13]. After estimating baseline SCr, we determined whether the patients had an AKI and subsequently classified them into AKI stages. A patient was defined as having AKI if one of the SCr measurements during the 7-day period was ≥ 150% of the baseline SCr or if the SCr increased ≥ 0.3 mg/dl within a 48-h interval during the 7-day period. Once AKI status was determined, AKI staging was performed. A 150–199%, 200–299%, and ≥ 300% increase from baseline SCr in one of the SCr measurements during the 7-day period was defined as stage 1, 2, and 3 AKI, respectively. Regardless of fulfilling these criteria, patients were defined as developing stage 3 AKI when SCr increased ≥ 4.0 mg/dl or RRT was initiated. We could not use the criteria for hourly urine output because data on hourly urine output were not collected in the registry.

Descriptive statistics are reported as median (interquartile range) or mean ± standard deviation for continuous variables according to the normality of distribution. Data normality was analysed using Shapiro–Wilk test or Kolmogorov–Smirnov test. Categorical variables are reported as frequency (percentage). Demographics and clinical differences between groups were assessed using Pearson’s chi-squared test, Fisher’s exact test, independent sample t-test, or Mann-Whitney U test, as appropriate. We performed multivariate Cox regression analysis to calculate the hazard ratio (HR) and 95% confidence interval (CI) of RRT with respect to mortality and to find independent factors associated with risk of death. All variables were entered into a multivariate Cox regression analysis. Differences in Kaplan-Meier survival curves between RRT and non-RRT groups were compared using the log-rank test. A P value < 01.05 was considered statistically significant. Statistical analyses were performed using IBM SPSS version 25.0 (IBM Corp., Armonk, NY, USA).

Out of 10,426 patients with OHCA that were screened during the study period, 1373 were treated with TTM at 22 academic hospitals throughout South Korea (Fig. 1). We excluded the patients who died within 48 h of ROSC to minimise competing risk, those with pre-arrest chronic kidney disease or end-stage renal disease, and those with incomplete data such as SCr, survival, or neurological status at 6 months. In total, 1063 patients were enrolled in the study cohort. The baseline characteristics of stage 3 patients with AKI according to RRT status and outcomes are summarised in Tables 1 and 2.

AKI was defined in 590/1063 (55.5%) of OHCA patients treated with TTM. Most AKI developed within 3 days of ROSC (461/590 [78.1%]). The frequency of AKI stage was 257/590 [44.5%], 110/590 [18.6%], and 223/590 [37.8%] for stages 1, 2, and 3, respectively. RRT was applied in 115/223 (51.6%) patients with stage 3 AKI. The most common modality among RRT patients was CRRT (111 [96.5%]), and the remaining four patients (3.5%) received intermittent haemodialysis. The average initiation time of RRT was 18 (9–46) h from the time that ROSC and RRT were initiated, which was within 48 h of ROSC in most cases (89/115 [77.4%]). The time to achieve AKI stage 3 in patients with RRT was 1 (1–2) day, and the time to achieve AKI stage 3 in patients without RRT was 3 (2–4) days. Average SCr values from day 1 to 7 following ROSC is provided in the supplementary material (see Additional file 1, Table S1).

The overall 6-month mortality of all enrolled patients (entire cohort, n = 1063) was 541/1063 (50.9%). The 6-month mortality of the non-RRT group in the entire cohort was significantly lower than that of the RRT group (448/948 [47.3%] vs. 93/115 [80.9%], P < 0.001) (Fig. 2). In case of patients with AKI (AKI cohort, n = 590), 6-month mortality was 380/590 (64.4%). The 6-month mortality of the non-RRT group in AKI cohort also was significantly lower than that of the RRT group (287/475 [60.4%] vs. 93/115 [80.9%], P < 0.001). However, in patients with stage 3 AKI (stage 3 AKI cohort, n = 223), the 6-month mortality of the non-RRT group was significantly higher than that of the RRT group (98/108 [90.7%] vs. 93/115 [80.9%], P = 0.04) (Table 2). The Kaplan-Meier survival curves differed significantly between the non-RRT and RRT groups in these cohorts (P = 0.003) (Fig. 3).

In the multivariate Cox regression analysis, four variables, namely, body weight, hypoglycaemia, re-arrest, and RRT, were independently associated with risk of death in patients with stage 3 AKI (Table 3). The HR of RRT was 0.569 (95% CI, 0.377–0.857, P = 0.01).

The distributions of CPC (1–5) at 6 months for the non-RRT vs. RRT groups were 3/108 (2.8%) vs. 12/115 (10.4%) for CPC 1, 0/108 (0.0%) vs. 1/115 (0.9%) for CPC 2, 1/108 (0.9%) vs. 3/115 (2.6%) for CPC 3, 6/108 (5.6%) vs. 6/115 (5.2%) for CPC 4, and 98/108 (90.7%) vs. 93/115 (80.9%) for CPC 5, respectively (P = 0.01) (Table 2).

Oliguria or anuria was the leading cause of initiating RRT (62/115 [53.9%]). Other causes followed acidosis (30/115 [26.1%]), azotaemia (9/115 [7.8%]), volume overload (8/115 [7.0%]), and hyperkalaemia (6/115 [5.2%]). Reasons for terminating RRT included death (67/115 [58.3%]), recovery from AKI (20/115 [17.4%]), withdrawal of RRT continuation by patient’s legal surrogates (18/115 [15.7%]), and discharge/transfer of the patient (10/115 [8.7%]).

Only 32 RRT patients survived to discharge (32/115 [27.8%]). Among them, 8/32 (25%) still needed RRT at discharge.

The present study has several limitations. First, because we defined AKI using only SCr criteria, it is possible that we underestimated AKI prevalence in our cohort. In the case of critically ill patients, the production of SCr decreases. Additionally, a decrease in SCr could be potentiated by therapeutic hypothermia [26]. Therefore, an exact judgement on AKI status is difficult in the case of individuals treated with therapeutic hypothermia. In addition, differences in urine output and fluid balance are likely to have influenced the decision to start RRT. However, data on urine output at RRT initiation and fluid balance could not be evaluated in our cohort. Second, the present study was a retrospective analysis of prospectively collected observational data that did not include information on standardised treatments or management protocols for RRT. Therefore, the risk of selection bias could have been introduced. Third, the initiation of RRT in our study was at the discretion of the physician responsible for the patient’s care according to the KDIGO guideline (e.g., severe hyperkalaemia, severe acidosis, pulmonary oedema, and uremic complication). There were no consistent protocols dictating when to initiate RRT, which might have led to inconsistencies in practices among the principal investigators of each institution. A randomised controlled trial of RRT on patients with stage 3 AKI is needed to confirm the role of RRT. Fourth, there was a possibility of failure to conduct RRT even if urgent RRT was indicated because of the decision on withdrawal of life-sustaining therapy according to the neurological prognostication. However, in South Korea, the withdrawal of life-sustaining therapy depending on the neurological prognostication has not been used widely because of several reasons. For instance, the life-sustaining treatment decisions act was only applied recently in South Korea (since February 2018). In addition, legal surrogates, including family members, could not determine withdrawal of life-sustaining therapy, especially stopping the ventilator or removing the endotracheal tube owing to traditional sentiment. Therefore, we expect that decisions regarding withdrawal of life-sustaining therapy may not affect the initiation of RRT. However, there was a possibility that RRT may be stopped when a poor neurological outcome or death was expected through the treatment course, as the cost for RRT (especially continuous RRT) is quite expensive in South Korea (approximately €400 per day). Fifth, in the cohort of RRT (n = 115), 80 patients were diagnosed as achieving stage 3 AKI by initiating RRT itself (Group A) and 35 patients were diagnosed as achieving stage 3 AKI based on SCr (Group B). Although the initiation of RRT was not determined by conditions other than SCr in most cases, daily SCr of Group B was significantly higher than that of Group A (see Additional file 1, Table S2). Therefore, it may be possible to compare patients reaching the AKI stage 3 by non-creatinine criteria with patients reaching the AKI stage 3 by creatinine criteria. The sixth limitation is that the results of the present study may have been affected by the characteristics of our study cohort. Therefore, the results might not be reproducible in another cohort with different characteristics. Fifth, although we excluded the patients who died within 48 h from the time of ROSC initiation to decrease problems related to competing risk, excluding these patients itself may have introduced bias.