Serum Lactate and Mortality during Pediatric Admissions: Is 2 Really the Magic Number?

Abstract

The primary objective of this study was to determine if serum lactate level at the time of hospital admission can predict mortality in pediatric patients. A systematic review was conducted to identify studies that assessed the utility of serum lactate at the time of admission to predict mortality in pediatric patients. The areas under the curve from the receiver operator curve analyses were utilized to determine the pooled area under the curve. Additionally, standardized mean difference was compared between those who survived to discharge and those who did not. A total of 12 studies with 2,099 patients were included. Out of these, 357 (17%) experienced mortality. The pooled area under the curve for all patients was 0.74 (0.67–0.80, p  < 0.01). The pooled analyses for all admissions were higher in those who experienced mortality (6.5 vs. 3.3 mmol/L) with a standardized mean difference of 2.60 (1.74–3.51, p  < 0.01). The pooled area under the curve for cardiac surgery patients was 0.63 (0.53–0.72, p  < 0.01). The levels for cardiac surgery patients were higher in those who experienced mortality (5.5 vs. 4.1 mmol/L) with a standardized mean difference of 1.80 (0.05–3.56, p  = 0.04). Serum lactate at the time of admission can be valuable in identifying pediatric patients at greater risk for inpatient mortality. This remained the case when only cardiac surgery patients were included.

Introduction

Lactate is a normal by-product of glucose and pyruvate metabolism. Under anaerobic conditions, pyruvate, the end-product of glycolysis, is converted to lactate by the enzyme lactate dehydrogenase (LDH). ¹ Lactate contributes to the production of cellular energy through several mechanisms such as the Cori cycle for gluconeogenesis and the production of nicotinamide adenine dinucleotide+ (NAD + ) for glycolysis. ² Traditionally, increased levels of serum lactate were only associated with hypoperfusion; however, more causes have been recognized and should be considered. ³ Elevated serum lactate levels can be caused by both an increase in its production and a decrease in its clearance. ⁴ The pathophysiology of hyperlactatemia has been a subject of debate, because it is not simply caused by anaerobic metabolism. ⁵ Lactate levels may increase due to various etiologies including increased anaerobic glycolysis, hyperadrenergic status, impaired hepatic clearance, inhibited pyruvate dehydrogenase, and mitochondrial dysfunction. ^{1 2 5 6}

Lactate was first proposed as a metric for the prediction of mortality in patients presenting with shock by Broder and Weil. ⁷ Since then, many studies have been published and have shown that elevated lactate levels are associated with morbidity and mortality, making its measurement useful for risk stratification. ^{7 8 9 10 11 12 13} Studies have demonstrated its utility even in the setting of organ dysfunction and shock. ^{14 15} Beyond the absolute value of lactate, studies demonstrate that more rapid lactate clearance is associated with decreased mortality and that lactate-guided therapy may be complementary to early goal-directed therapy in reducing mortality. ^{16 17 18} Serial measurement of lactate in adults has shown utility as decreasing levels are associated with a better prognosis in all common situations of hyperlactatemia regardless of the initial value. ⁵ However, this does not hold true in every scenario. ¹⁹

In children, the use of serum lactate to predict mortality has not been delineated in as much detail as it has in adults. However, the physiology of lactate production and clearance is not dramatically difference between children and adults and there is little reason to believe that there should be a large difference in the utility of lactate in children compared with adults. The Surviving Sepsis Campaign guidelines were unable to issue a recommendation regarding the measurement of serum lactate to stratify children with suspected septic shock. These guidelines, do, however, recommend its measurement to guide resuscitation as normalization of its levels is associated with decreased persistent organ dysfunction. ^{20 21} A serum lactate greater than 2 mmol/L is often used as a clinical and academic cutoff for lactate levels. ^{9 20 22} This cutoff, however, is based on little objective evidence. Outside of septic shock, pediatric cardiac surgery is a particular area of interest for early detection of change in hemodynamic status. The use of lactate levels at admission as a tool for prediction of mortality has been subject of debate. ²³ Therefore, the primary objective of this study was to determine if serum lactate level at the time of hospital admission can predict mortality in pediatric patients and the secondary objective was to determine its value in pediatric patients that have undergone cardiac surgery.

Materials and Methods

This study did not require institutional review board approval as they used previously published data that were deidentified. This study follows the Helsinki declaration.

Manuscript Search and Identification Strategy

A systematic review of the literature was performed to identify manuscripts describing comparisons of serum lactate at the time of hospital admissions between those who survived to discharge and those who did not. The primary variable of interest was serum lactate at the time of admission in pediatric patients. The reporting of this systematic review was guided by the standards of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) Statement. ²⁴

Manuscripts were identified using electronic databases including PubMed, EMBASE, Ovid, and Cochrane reviews. These databases were queried using the following search terms individually and in various combinations: “serum,” “lactate,” “lactic acid,” “pediatric,” “children,” “survival,” and “mortality.” Only studies in English language were eligible for inclusion. No specific restriction on year of publication was used. The final search was conducted on November 1st, 2020. Resulting studies were screened by title and abstract. Those felt to be pertinent to help fulfill the objectives then had their full text retrieved in their entirety. References of these studies were then hand searched for additional relevant manuscripts. No direct contact with manuscript authors was made to obtain full text manuscripts or additional data.

The full-text manuscripts were then reviewed and assessed for quality. Published manuscripts available in full text were included in this review if they presented data comparing serum lactate on admission in pediatric patients who survived to discharge and those who did not survive to discharge.

Inclusion Criteria

The following inclusion criteria must have been met for a study to be included in the study: (1) must have had pediatric patients only (less than 18 years of age); (2) studies must have reported serum lactate values on admission; (3) serum lactate levels must have been compared between those who survived to discharge to those who did not; (4) a receiver operator curve analysis must have been conducted by the study to determine a cutoff value for serum lactate to predict mortality; (5) serum lactate cutoff point and area under the curve (AUC) must have been reported for the receiver operator curve analyses.

Study identification was conducted separately by two authors (RL and SF). Studies identified for inclusion by these two authors were then reviewed by a third author (EV). Any discrepancies between the two authors were identified by the third author and reviewed by all authors to come to a consensus.

Data Extraction

Data regarding baseline patient characteristics and study characteristics were extracted from the manuscripts identified for inclusion. Study level data were extracted with the use of a data collection form that was developed specifically for this review. The data extraction was conducted by two separate authors (RL, EV) to ensure integrity of the resulting data. Differences were then identified by a third author (JF). Discrepancies in the data extraction between the two authors were then reviewed by all authors to come to a consensus. With respect to the serum lactate value, the admission value was the only one collected. For studies with patients undergoing cardiac surgery, the serum lactate at the time of admission to the intensive care unit was extracted. For a single study, the value used was a serum lactate drawn in the operating room after cardiopulmonary bypass was ceased and 20 minutes after protamine administration. This serum lactate value is not expected to be significantly different from the serum lactate drawn at the time of admission to the intensive care unit and thus was included.

Study Quality Assessment

Bias in the included studies was assessed using the Newcastle-Ottawa Scale (NOS). ²⁵

Data Analysis

Continuous data are presented as mean and standard deviation. Categorical data are presented as frequencies with absolute numbers as well as percentages. Pooled analyses were conducted using MedCalc Version 19.2.6 (MedCalc Software Ltd, Ostend, Belgium).

The first set of pooled analyses were done to pool AUC for receiver operator curve analyses to determine the accuracy of serum lactate on hospital admission to predict mortality. One pooled analysis was done including all the identified studies, while a second analysis was done only including the identified studies in which children were admitted to an intensive care unit after cardiac surgery. Heterogeneity was assessed and if heterogeneity was found to be significant, then a random-effects model was used, otherwise a fixed-effects model was used.

The second set of pooled analyses were done to compare serum lactate levels in those who survived to discharge to those who did not. This was done utilizing the mean and standard deviation of the study-level data. If individual studies presented their data as median and range, the mean and standard deviation were calculated. One pooled analysis was done including all the identified studies, while a second analysis was done only including identified studies in which children were admitted to an intensive care unit after cardiac surgery. Heterogeneity was assessed and if heterogeneity was found to be significant, then a random-effects model was used, otherwise a fixed-effects model was used.

Publication Bias Analysis

Publication bias was assessed by visual assessment of the publication bias funnel plot.

Results

Study Characteristics

Initial search conducted as outlined above yielded 1,014 manuscripts after duplicates were removed. After reviewing study titles and abstracts, full text was obtained for 36 studies. Of these 36 studies, a total of 12 with 2,099 patients were included in the final analyses ( Fig. 1 ). ^{23 26 27 28 29 30 31 32 33 34 35 36} Of these 12 studies, 4 (33%) included patients after cardiac surgery, 3 (25%) included patients with shock, and the remainder included patients with various diagnoses. Out of the 2,099 patients, 357 (17%) experienced mortality. The mean age was 2.1 years ( Table 1 ).

Fig. 1.

Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) flowchart.

Table 1. Characteristics of included studies.

Author	Year	Study design	Patient population	Patients ( n )	Mortality ( n )	Survivor ( n )	Mean age (years)	NOS score
Alam and Gupta 26	2020	Prospective single-center cohort	Shock	116	58	58	2.5	6
Bai et al 27	2014	Prospective single-center cohort	PICU	1109	115	994	1	7
Cheung et al 23	2005	Prospective multicenter cohort	Cardiac surgery	85	14	71		7
Durward et al 28	2005	Prospective single-center cohort	Cardiac surgery	85	5	80	0.4	7
El-Mekkawy et al 29	2020	Prospective single-center cohort	PICU	78	13	65	1.5	8
Gupta et al 30	2020	Prospective single-center cohort	PICU	78	16	62	2.8	7
Haidar et al 31	2020	Retrospective single-center cohort	Acetaminophen-induced liver failure	95	49	46	1.1	6
Hatherill et al 32	2003	Prospective single-center cohort	Shock	46	16	30	0.5	7
Kapoor et al 33	2016	Prospective single-center cohort	Cardiac surgery	150	11	139		8
Lee et al 34	2020	Retrospective single-center cohort	Shock	26	14	12	12	7
Ma et al 35	2019	Prospective multicenter cohort	Pneumonia	155	22	133	11	7
Rocha et al 36	2010	Retrospective single-center cohort	Cardiac surgery	76	24	52	0.03	8

Abbreviations: NOS, Newcastle-Ottawa Scale; PICU, pediatric intensive care unit.

Pooled Analyses of Serum Lactate Levels on Admission, All Patients

A total of 12 studies with 2,099 patients were included in this analysis. The Q-statistic for heterogeneity had a p -Value of less than 0.01 and the I-squared value was 96%, demonstrating significant heterogeneity. Thus, a random-effects model was used. Serum lactate levels at the time of admission were higher in those who experienced mortality (6.5 vs. 3.3 mmol/L). This resulted in a standardized mean difference of 2.60 (95% confidence interval: 1.74–3.51, p -Value of less than 0.01) ( Fig. 2 ).

Fig. 2.

Forest plot showing pooled analyses of receiver operator curves in all patients.

Pooled Analyses of Receiver Operator Curves, All Patients

A total of 12 studies with 2,099 patients were included in this analysis. The Q-statistic for heterogeneity had a p -Value of less than 0.01 and the I-squared value was 93%, demonstrating significant heterogeneity in this endpoint. Thus, a random-effects model was used. The pooled AUC for serum lactate at the time of admission was 0.74 (95% confidence interval: 0.67–0.80, p -Value of less than 0.01). This demonstrates that serum lactate at the time of admission had acceptable value to predict mortality ( Fig. 3 ).

Fig. 3.

Forest plot showing pooled analyses of receiver operating characteristic curves (ROC) in cardiac surgery patients.

Pooled Analyses of Serum Lactate Levels on Admission, Cardiac Surgery Patients

A total of four studies with 396 patients were included in this analysis. The Q-statistic for heterogeneity had a p -Value of less than 0.01 and the I-squared value was 96%, demonstrating significant heterogeneity. Thus, a random-effects model was used. Serum lactate levels at the time of admissions were higher in those who experienced mortality (5.5 vs. 4.1 mmol/L). This resulted in a standardized mean difference of 1.80 (95% confidence interval: 0.05–3.56, p -Value of 0.04) ( Fig. 4 ).

Fig. 4.

Forest plot showing pooled analyses of serum lactate levels on admission in all patients.

Pooled Analyses of Receiver Operator Curves, Cardiac Surgery Patients

A total of four studies with 396 patients were included in this analysis. The Q-statistic for heterogeneity had a p -Value of less than 0.01 and the I-squared value was 83%, demonstrating significant heterogeneity in this endpoint. Thus, a random-effects model was used. The pooled AUC for serum lactate on admission was 0.63 (95% confidence interval: 0.53–0.72, p -Value of less than 0.01). This demonstrates that serum lactate at the time of admission had poor value to predict mortality ( Fig. 5 ).

Fig. 5.

Forest plot showing pooled analyses of serum lactate levels on admission in cardiac surgery patients. ROC, receiver operating characteristic.

Discussion

These pooled analyses demonstrate that serum lactate at the time of pediatric admissions has acceptable value to predict mortality when all pediatric patients are considered. However, these analyses identified only fair value in serum lactate to predict mortality in pediatric patients after cardiac surgery. The mean serum lactate level at the time of admission for all pediatric patients who experienced mortality was 6.5 mmol/L, while the mean serum lactate level at the time of admission for those who underwent cardiac surgery was 5.5 mmol/L. Although previous studies have demonstrated these findings, the findings from these pooled analyses offer a quantitative summary.

Identifying patients who are more critically ill, and thus at greater risk for mortality, can be particularly helpful in clinical management, prognostication, and risk stratification. The predictive value, however, is lower in those admitted after cardiac surgery. Serum lactate is, anecdotally, among the most widely and frequently used of these. Many utilize a cutoff of 2 mmol/L for concern, although the cutoff itself seems to have unfounded origin. Data from these pooled analyses, in fact, demonstrate that a cutoff of 2 mmol/L in fact is well under the cutoff that appears to predict mortality.

Lactate is produced by most tissues in the body, although muscle is the primary location. ^{3 37} It is produced from pyruvate, which is the end-product of glycolysis. Under aerobic conditions, pyruvate enters the Krebs cycle bypassing lactate production. Nevertheless, under anaerobic conditions, LDH converts it to lactate in the cytoplasm. Lactate is used to produce cellular energy by different mechanisms. ² First, its production generates NAD ⁺ , which is used for glycolysis; second, it is a substrate for gluconeogenesis via the Cori cycle in the liver ³ ; and last, it serves as an important energy substrate in the brain through the cell-to-cell shuttle. ²

As any other biomarker, hyperlactatemia occurs when its production exceeds its clearance and can be found in different clinical settings. ⁴ Most characteristically, it is found in a state of global hypoperfusion. Most of the literature published is on septic shock, but any form of shock or tissue hypoperfusion can lead to an increase in lactate. ³ However, it can also occur in other situations such as seizures, heavy exercise, use of certain drugs, excessive work of breathing, thiamine deficiency, and liver failure. ³

Serum lactate increases due to anaerobic glycolysis in hypoperfused tissue, an hyperadrenergic state, impaired hepatic clearance, inhibition of pyruvate dehydrogenase, and mitochondrial dysfunction. ^{1 2 6} Due to lactate being a normal product of glucose and pyruvate metabolism, any increases in glycolysis or decreases in pyruvate metabolism will cause its levels to rise; hypoperfusion secondary to any cause may result in this. ² In a state of hypoperfusion, glycolysis will increase. First, due to the augmented need for anaerobic metabolism; and to support an increase in glycolysis, NAD ⁺ is required from the conversion of pyruvate to lactate by LDH, thus contributing to the rise of lactate levels. ¹ Second, glycolysis will increase due to a hyperadenergic state, which arises in compensation to hypoperfusion. Accordingly, an increase in catecholamines will further promote glycolysis due to stimulation of the β ₂ -adrenergic receptors. ³⁸

Another important contributor to hyperlactatemia is impaired clearance. Lactate is mostly cleared by the liver; however, the kidneys contribute with a small percentage. ³⁹ Decrease in hepatic clearance may be mediated by a variety of factors, not only due to liver hypoperfusion. ^{4 40 41} In certain clinical settings, hepatic gluconeogenesis is also affected, and glycolysis becomes the primary mode of energy production in the liver; therefore, the liver becomes a lactate producing organ. ¹ Moreover, acute or chronic liver failure further exacerbates hyperlactatemia. ^{42 43 44} In addition, it has been reported that patients with sepsis have limited pyruvate metabolism due to the reduced function of pyruvate dehydrogenase, which converts pyruvate to acetyl-coenzyme A, further contributing to the formation of lactate. ⁴⁵

As mentioned previously, the mechanisms in which serum lactate increases, particularly in sepsis, are complex and could be challenging to interpret. Clinicians should not only consider hypoperfusion as the only mechanism behind persistent increased lactate but also consider the hyperadrenergic status and an impaired lactate clearance, as pursuing aggressive fluid resuscitation in patients whose cause of elevated lactate is not hypoperfusion may lead to fluid overload and worsened outcomes. ⁴⁶ Therefore, it is important for clinicians to keep in mind that a single laboratory value should not guide therapy and the assessment of other parameters such as central venous pressure, capillary refill time, and urine output is needed.

Nevertheless, this analysis is aligned with the majority of evidence from adult patients; in that elevated lactate levels are associated with an increased risk of mortality. ^{8 9 10 11 12 13 14} However, this does not automatically mean that the resuscitative efforts should be more aggressive. Various authors have suggested that the interpretation of lactate levels in adults should be separate from the decisions to guide resuscitative efforts and that lactate clearance should not be the “end point” of resuscitation. ^{47 48} Still, as this analysis shows, serum lactate levels can be valuable as a prognostic tool to identify greater risk of mortality in pediatric patients.

Particularly in pediatric patients admitted after cardiac surgery, the value of a single lactate has been widely debated with studies showing mixed results. One study by Siegel et al showed a mean lactate level on admission of 2.38 mmol/L for survivors and of 6.86 mmol/L for nonsurvivors; a lactate level of 4.5 mmol/L had an 100% positive predictive value (PPV). ⁴⁹ Other studies have supported these findings. ^{50 51 52} However, a study by Hatherhill et al found that a lactate value of 6 mmol/L at admission predicted death with a sensitivity of 78%, a specificity of 83%, and a PPV of only 32% and concluded that its routine use as a predictor of mortality could not be justified. ⁵³ While this analysis found that serum lactate levels at the time of admission in patients after cardiac surgery were higher in those who experienced mortality (5.5 vs. 4.1 mmol/L), it had poor value to predict mortality with an AUC of 0.63 (95% confidence interval: 0.53–0.72).

Some authors have suggested that serial lactate levels would be a better prognostic tool. Charpie et al studied neonates after complex congenital heart surgery and showed that a change in lactate level of 0.75 mmol/L per hour or more predicted a poor outcome with an 89% sensitivity value, a 100% specificity value, and a 100% PPV. ⁵⁴ Another study showed that Lactime, time during which the lactate remains greater than 2 mmol/L, of 48 hours or more has a PPV of 60% and concluded that it is a useful predictor of mortality in children undergoing repair or palliation of congenital heart disease. ⁵⁵ Additionally, Schumacher et al showed that a postoperative lactate increase rate of 0.6 mmol/L/h allows for discrimination between high and low risk of mortality after in neonates of both single and biventricular physiology after undergoing congenital heart surgery. ⁵⁶ It is important to keep in mind that lactate levels in patients that have undergone cardiac surgery can be elevated by other causes besides global tissue hypoxia; other variables that could influence the lactate levels in the postoperative period are the duration of total circulatory arrest, the depth of intraoperative anesthesia, and the use of some adjunctive medications. ⁵³

While the aforementioned studies demonstrate some variable findings, the current pooled analyses do demonstrate statistically significant differences in admission lactate levels and outcomes. In fact, this variability in findings highlights the need for such pooled analyses. While AUC analyses could not be done, the current findings indicate that lactate can be helpful as a prognostic tool.

These analyses, while additive, are not without their limitations. First, data are all study-level data and not patient-level data. Thus, patient-specific confounders cannot be accounted for, simply only study-specific confounders. Second, these pooled analyses do not allow for robust adjustment for confounders for the previously mentioned reason in conjunction with the relatively low number of studies included in the pooled analyses. Additionally, these data come from patients admitted for various clinical indications. This was attempted to be overcome by providing a subset analysis for those undergoing cardiac surgery as data was available from a reasonable number of studies for this. Significant heterogeneity was present in all the pooled analyses but this in and of itself is not a limitation. In fact, the presence of heterogeneity speaks to the need for such pooled analyses as a random-effects model helps provide a meaningful, quantitative summary of otherwise heterogeneous data. While the current pooled analyses do not assess the value of serum lactate in combination with other clinical parameters, serum lactate in combination with other clinical parameters can have even greater value in such risk stratification of patients. Selection bias may be present in the analyses as many studies utilize data from patients with sepsis. This is likely due to the fact that, historically, this a clinical setting in which lactates have been frequently used. Nonetheless, by including all pediatric data that is available, we are hopefully this is overcome to some extent. Interestingly, findings between the two most frequent clinical states were not remarkably different. Some bias may also be introduced by studies from regions with different resources. The study by Alam and Gupta is from a lower resource setting and did have a higher mortality than other studies. The lactate values themselves did not seem vastly different between the two groups. Nonetheless, this does introduce some resource-based bias, Lastly, we only included studies published in English. This was done to ensure that the included papers were able to be properly understood by all the authors. Not all the authors are proficient in similar other languages outside of English and thus non-English papers were excluded.

Conclusion

Serum lactate at the time of admission can be valuable in helping identify pediatric patients at greater risk for inpatient mortality. This remained the case when only cardiac surgery patients were included. Identification of higher risk patients may allow for more appropriate clinical resource allocation.

Funding Statement

Funding None.