Blood Lactate or Lactate Clearance: Which Is Robust to Predict the Neurological Outcomes after Cardiac Arrest? A Systematic Review and Meta-Analysis

Aims Lactate and lactate clearance were supposed to be associated with cardiac arrest outcomes, but studies obtained different results. Thus, we conducted this meta-analysis to investigate the association between lactate or lactate clearance and neurological outcomes and their usefulness for prediction of neurological outcomes. Methods We conducted a systematic search in PubMed, Web of science, EMBASE, Medline, and Google Scholar until May 1, 2018, for relevant studies. Studies reporting lactate, lactate clearance on admission, or other time points after admission associated with neurological outcomes were included in our analysis. Pooled effect date was shown as weighed mean difference (WMD) and 95% confidence interval (CI). To measure the usefulness of lactate on admission to predict neurological outcomes, we also pooled the data of diagnostic test. Results 23 studies involving 6720 cardiac arrest (CA) patients were included. Results from our analysis indicated that patients with good neurological outcomes tended to have a lower lactate level on admission (WMD: -2.66 mmol/L, 95%CI: -3.39 to -1.93) and 12h, 24h, and 48h after admission (P<0.001). Furthermore, the pooled AUC for lactate level on admission to predict neurological outcomes was 0.77 (95%CI: 0.73-0.80). However, a significant association between lactate clearance and neurological outcomes was only found in 24h but not 12h lactate clearance rate. Conclusions Lactate levels on admission and all time points up to 48h were associated with neurological outcomes after CA, whereas the association between lactate clearance and neurological outcomes was not so stable. Lactate was a more robust surrogate marker than lactate clearance to predict neurological outcomes after CA.


Introduction
Cardiac arrest (CA) presents a serious public health concern with high mortality cause by various pathogenesis such as anoxia, drug, electrolyte disturbance, low temperature, and hypoglycemia [1]. In Europe and north American, CA causes more than 600,000 cases per year [2], and the survival rate at hospital discharge is very low. Two major reasons for the mortality after resuscitation are postresuscitation circulatory failure and postanoxic neuro injury. Many researches have performed studies to develop tools for predicting the neurological outcomes. Biological markers such as neuron-specific enolase and S-100 protein have been demonstrated to be efficient predictors for neurological outcomes after CA. But these markers need longer time for detection. Therefore, a facility marker to predict neurological is needed in clinical practice. Lactate and lactated clearance are supposed to be that kind of marker.
Lactate is a product of pyruvate reduction during glycolysis. During CA, tissue hypoxia leads to accumulation of pyruvate and therewith the accumulation of lactate. The level of circulation lactate is likely to be a good marker of tissue hypoxia which is also recommended by Utstein guidelines. Not only lactated level but also a lactate clearance 2 BioMed Research International from serial determination of lactate is supposed to associated with cardiac arrest outcomes of CA [3,4], and a serial determination of lactate is also recommended in CA survival after return of spontaneous circulation (ROSC) [5,6]. But studies obtain discrepant results [3,4,[7][8][9][10][11]. This also leads to some confusions in the application of lactate or lactate clearance to predict neurological outcomes in clinic. In order to figure out the relationship between lactate or lactate clearance at different time points and neurological outcomes after CA and compare the efficiency of lactate or lactate clearance in predicting neurological outcomes after CA, we carry out a meta-analysis to summarize the current evidence. Besides, we also reviewed the reference lists of relevant articles to identify additional studies. Letters to author or abstract were scrutinized and excluded because they lacked sufficient data for analysis.

Materials and Methods
. . Study Selection. Studies were selected into our metaanalysis based on the following criteria: (1) carried on nontrauma cardiac arrest patients; (2) reported lactate level on admission or other time points after admission related to neurological outcomes; (3) reported lactate clearance related to neurological outcomes; (4) reported the detailed data of prognostic test. Studies were excluded in the following cases: (1) if it did not report lactate level related to neurological outcomes, (2) letter to author, and (3) study abstract.
. . Data Extraction. The following data were extracted from each eligible study: the first author's last name, year of publication, study design, geographic location, number of subjects, mean age of subjects, gender of subjects, percentage of initial shockable rhythm, location of CA, outcome followup time, therapeutic hypothermia, the sources where lactate measures from, neurological outcomes, and mean lactate level on admission. Besides, the data of serum lactate level and lactate clearance related to neurological outcomes at different time points were also extracted. To make an overview of lactate diagnostic efficiency, date of diagnostic test was also extracted. Lactate cut-off point, lactate measured time, neurological outcome measured time, and absolute value of true positive, true negative, false positive, and false negative were retrieved or developed 2x2 contingency table. Study selection and date extraction were conducted independently by two authors (Z-BC and Z-Z), with any disagreements resolved by consensus.
. . Methodological Quality Assessment. The methodological quality assessment of eligible studies was appraised using Newcastle-Ottawa Scale (NOS) [32]. Articles with scores <4, between 4 and 6, or >6 were considered as low, intermediate, or high quality, respectively. The assessment was also carried out by two authors (Z-BC and Z-Z). If there was any disagreement, a third author would reevaluate the original study.
. . Statistical Methods. Effect size of lactate difference between different neurological outcomes patients was defined as weighed mean difference (WMD) and 95% confidence interval (CI). Articles reported data as interquartile range and median [4, 10, 13-17, 19, 21, 23-25, 27, 30, 31]. We converted data into mean and s.d. according to the method developed by Wan [33]. The unit of lactate was uniformed to mmol/L. Lactate measured at initial, on the arrival of hospital, emergency department, or ICU was defined as admission lactate level. Lactate measured within one hour after admission was also defined as lactate on admission. Besides, we also analyzed lactate level at 12h, 24h, and 48h after admission related to neurological outcomes. We used the Cochran Q and I 2 statistics to assess heterogeneity among studies. For the Q statistic, a P value < 0.1 was considered statistically significant heterogeneity. For the I 2 statistics, a value greater than 50% was considered high heterogeneity [34]. We used a random-effects model to estimate WMD in case of heterogeneity. Statistical synthesis and data analysis of diagnostic test were conducted according to the method introduced by Lee [35]. Sensitivity, specificity, likelihood ratios, diagnostic odds ratios, and receiver operating characteristic curves (SROC) were pooled using the DerSimonian and Laird method (random-effects model). To analyze the threshold effect, Spearman correlation coefficient of sensitivity and 1-specificity was calculated.
Considering that the differences in trails design and baseline characteristic of patients may contribute to the obvious heterogeneity observed in our analysis, we also designed subgroup analysis and metaregression analysis to explore the vital baseline factors. We carried out subgroup analysis according to mean age (adult versus old), CA location (  transformation (Yes versus No), and outcome follow-up time (long versus short-term). Old people were defined as patients above 65 years old. Long-term follow-up was defined as measured neurological outcomes 3 months or longer after CA. Metaregression analysis carried out subgroup analysis according to mean age, percentage of initial shockable rhythm, and mean lactate level on admission. Sensitivity analysis was employed to assess the stability of the meta-analysis. The publication bias was assessed using Egger [36] and Begg [37] tests, and P value less than 0.05 was considered to have a significant publication bias. All statistical analyses were performed using the STATA version 14.0 and Meta-DiSc version 1.4. Except where otherwise specified, a P value <0⋅05 was considered to be statistically significant.

Results
. . Literature Search. The systematic literature search yielded 382 potentially relevant records ( Figure 1). After excluding duplicates and clearly irrelevant publications by reading titles and abstracts, we obtained 41 full articles of potentially relevant studies for a further evaluation. After full-text reviews, 18 out of 41 articles were excluded. Six articles reported lactate levels related to survival outcome; 4 articles did not provide sufficient data; 2 articles carried on CA patients under the age of 18; 2 articles included trauma cardiac arrest patients; 2 articles were letter to author; 1 was system review; and 1 was study abstract. Finally, 23 articles fulfilled our inclusion criteria [3, 4, 10, 12-31].
. . Sensitivity and Publication Bias Analysis. In sensitivity analysis, there was not one study substantiality that subverted the WMD of lactate between different neurological outcomes except lactate level at 48h ( Figure S2). The funnel plots showed no obvious dissymmetry in the shape of funnel plots in both time points ( Figure S3). The P value of Begg rank correlation test and Egger liner regression test for lactate on admission was 0.463 and 0.346, respectively. In other time points, the results of Begg and Egger test also suggested no evidence of publication bias (P > 0.05).

Discussion
To the best of our knowledge, this was the first meta-analysis paying a close attention to the association and prognostic value of blood lactate and lactate clearance on CA neurological outcomes through systematically reviewing 23 relevant  studies. We found a significant relationship between lactate levels on admission and 12h, 24h, and 24h after admission were associated with neurological outcomes. Particularly lactate level on admission has the diagnostic ability to predict neurological outcomes after CA. But a significant association between lactate clearance and neurological outcomes was only found in 24h but not 12h.
As an organ dysfunction metabolite, lactate level is related to prehospital care factors such as quality of bystander CPR, the duration of down time [38], and initial rhythm and post-CA care factors such as hypothermia therapy. Therefore lactate was measured routinely in cardiac arrest patients or even other critically ill patients to assess short or long-term prognosis [39]. In cardiac arrest patients, patients with a lower lactate level on admission have higher possibility of ROSC [40]. One recent meta-analysis by Debaty et al. [41] also revealed that patients with a lower serum lactate level on hospital admission were associated with better survival outcomes. In our case, we further explored the association between lactate or lactated clearance and neurological outcomes. And we successfully demonstrated the associations between lactate levels on admission and 12h, 24h, and 48h after admission and CA neurological outcomes. Additionally, we also found a robust prognostic value of lactate level on admission to predict neurological outcomes. But if we wanted to apply it as gold standard in post-CA care, it was necessary to combine lactate level with other markers (e.g., blood ammonia, vasopressor) [15,42]. Studies reported that serum lactate level dramatically decreased within early time after admission (0-6h) but then slowly decreased [3,4,28]. A prospective designed multicenter study with 543 OHCA patients also documented that, despite the initial lactate level, effective lactate reduction over the first 6h of postcardiac arrest care was associated with survival and good neurologic outcomes [43]. It was the early lactate clearance which dominated the decrease of serum lactate. In our study, there was no significant difference in 12h lactate clearance rate between good and poor   Figure 5: Summary weighed mean difference (WMD) and 95% confidence intervals for lactate clearance rate at 12h (a) and 24h (b) after admission between good and poor neurological outcomes. Table 2: Subgroup analysis of lactate on admission and neurological outcomes.