Does lactate measurement performed on admission predict clinical outcome on the intensive care unit? A concise systematic review

Abstract

Background

There is a need for practical, efficient and effective prognostic markers for patients admitted to the intensive care unit (ICU) with sepsis, to identify patients at highest risk and guide and monitor treatment. Although many biomarkers and scoring systems have been advocated, none have yet achieved this elusive combination. Most ICUs already use blood lactate concentrations to monitor patients but the evidence base for this application is unclear.

Methods

A systematic review of the last five years of evidence of effectiveness of lactate measurement in prediction of outcome in ICUs was performed.

Results

It was found that there is a lack of high-quality evidence, and no specific studies of prognostic accuracy. d- or l-Lactate concentrations measured in plasma, serum, whole blood or colonic washings were raised at admission in almost all patient groups, and were higher in patient groups who had the worst outcomes (in-hospital mortality, sequential organ failure). However, there was significant overlap in individual concentrations measured in those who died within 28 days of admission, or who developed multiple organ failure, and those who did not. For serum l-lactate concentrations, no specific cut-off value capable of predicting in-hospital mortality or sequential organ failure could be recommended.

Conclusions

The evidence reviewed suggested that whole blood, plasma or serum lactate measurement could not provide specific prognostic information for individual patients. There may be a role for monitoring for normalization of serum d- or l-lactate concentrations during goal-directed therapy in the ICU but further good-quality studies are needed. Measurement of the d-lactate stereoisomer shows promise, such that further studies are warranted.

Introduction

There is a need for effective prognostic markers for patients admitted to the intensive care unit (ICU) in order that appropriately aggressive therapy can be directed towards those that need it most. However, it remains the case that no single marker or combination has so far achieved the elusive requirement of clinical reliability, simplicity in everyday use and practicality.

There are a number of clinicopathological prognostic scoring systems such as the Acute Physiology and Chronic Health Evaluation score (APACHE), the Sequential Organ Failure score (SOFA) and the Simplified Acute Physiology score (SAPS). Each of these uses a complex scoring system involving clinical observations and physiological and laboratory measurements to construct a specific risk score for the individual patient. Prognostic scoring systems are complex and time-consuming to construct and require the use of specific software. Another drawback of some of the scoring systems (e.g. APACHE III) is that in more recent iterations, a license fee may be payable to the APACHE III license holder for the right to access and use the derived equation for calculation. In these circumstances, they therefore quickly become prohibitively costly for everyday use.

In addition to scoring systems, a plethora of biomarkers has been proposed, including cytokines, inflammatory proteins, serine proteinases, antiproteinases, matrix metalloproteinases, fibrinolytic enzymes and peptides and thrombolytic markers. All of those that have been evaluated so far have been either ineffective in predicting severity of outcome, or unable to be used in real time because methods of measuring them are not readily available on main laboratory platforms. More recently, a potential role for procalcitonin has been suggested, though recent evidence suggests that this marker may be best used in guiding antibiotic therapy rather than in the diagnosis of bacterial sepsis or prognostication.

There is a widely held view among intensivists in Britain and elsewhere, e.g. the Surviving Sepsis Campaign (http://ssc.sccm.org/bundles/individual_changes/serum_lactate) that admission and serial lactate measurements are useful biomarkers in this setting (although there is no consensus on the precise cut-off points that should be applied). There is also widespread availability of both laboratory-based and point-of-care analysis of this analyte. Thus, it was felt that a systematic evaluation of the evidence supporting clinical effectiveness of lactate measurement when used in this way was timely.

Objectives

There were two main objectives of this study:

To seek out published evidence of whether measurement of lactate at the time of admission in patients in intensive care settings is effective in predicting clinical outcome compared with clinical acumen, clinico-pathological scoring systems or other biochemical markers;

To prepare a concise systematic review of the evidence.

Methods

Criteria for selecting studies for this review

The following types of studies were deemed acceptable: randomized controlled trials (RCTs), case-control cohorts and observational cohorts. The last five years of publications were considered to ensure the review was up to date and to avoid problems due to rapidly changing ICU practices and procedures. Studies published between 9 September 2005 and 9 September 2010 were included in the critical evaluations.

Studies with the following types of participants were accepted: those involving adult humans (we concluded that a paediatric review is warranted but should be conducted separately), studies with participants between 18 and 75 years of age and those studies with cohorts that included both male and female participants.

Measurement of d- and/or l-lactate concentrations in physiological fluids was taken as the index test. Measurement of d- and/or l-lactate concentrations could be performed in any physiological fluid including blood, broncho-alveolar lavage fluid, thoracotomy tap fluid, urine, colonic washout fluids and faeces.

Comparisons were against the following reference tests: APACHE scoring, SOFA scores, SAPS scores, arterial pH and arterial PaO₂.

The primary outcome measure was 28-day in-hospital mortality. The secondary outcome measure was multiple organ dysfunction syndrome.

Sources of heterogeneity included variable delay to first measurement and variations in severity of disease and case-mix, and each of these factors was considered during the critical appraisal process.

Search methods for identification of studies

The PubMed, MEDLINE and Embase Databases were searched on 15 September 2010 using the search criteria in Appendix A (please see http://acb.rsmjournals.com/lookup/suppl/doi:10.1258/acb.2011.011227/-/DC1). ‘Grey’ literature (sales brochures, internal reports, technical briefs and other documents written and published by government departments, quasi-autonomous non-governmental organizations, educational institutions, commercial companies or other groups that are not distributed by typical publication routes and are not indexed by online publication databases) or theses were not reviewed as this may have introduced a potential source of bias and it was felt that this omission was unlikely to influence the outcome of this systematic review.

Data collection and analysis

All abstracts were reviewed for relevance by the group. Disagreements were resolved by discussion. When the group could not reach a consensus, one person was nominated (CC) to be the final arbiter. There was no meta-analysis.

The full text of all papers identified as relevant was formally reviewed using standard published methods. The quality of the papers was assessed using the Cochrane risk of bias tool (http://www.ohg.cochrane.org/forms/Risk%20of%20bias%20assessment%20tool.pdf). The presumption is that the lower the risk of bias, the higher the quality. Although this tool is designed for interventional rather than diagnostic studies, the number of questions deemed valid and relevant to the present review was sufficient. The full text of all relevant papers was further assessed using the CASP critical appraisal checklist for prognostic studies (http://www.phru.nhs.uk/pages/phd/resources.htm). The QUADAS tool was considered for use, but it was rejected as it represented a duplication of the CASP tool and did not explicitly address prognostic studies (http://www.biomedcentral.com/1471-2288/3/25). A record was made if diagnostic accuracy data were available: specificity, sensitivity, odds ratio, risk ratio, likelihood ratio and receiver operator characteristic curve (ROCC) analysis. The systematic review was reported against the PRISMA checklist for reporting (http://www.prisma-statement.org/statement.htm).

Results

Using the search criteria described in Appendix A, a total of 39 studies ^1–39 were identified. The number of studies found and the type of study is summarized in Figure 1.

Figure 1

Summary of literature include in the review

After review of the abstracts for each study, 11 studies ^{1,2,7,12,17,21,24,29,34–36} were immediately rejected as they were deemed not relevant. The citations for these papers are given in the references, and they are clearly identified. Full text papers of the remaining 28 studies ^{3–6,8–11,13–16,18–20,22,23,25–28,30–33,37–39} were obtained and reviewed as described. The results of the CASP and Cochrane Risk of Bias appraisals for each study are available in Appendix B, as an online supplementary file (please see http://acb.rsmjournals.com/lookup/suppl/doi:10.1258/acb.2011.011227/-/DC1). (It is important to note that papers were reviewed in respect of the specific purpose of the present review, which in most cases was not the study's original stated purpose.) A brief overview of study results follows.

Types of study

There were two RCTs that evaluated diagnostic or prognostic accuracy of lactate measurement. ^11,30 The majority ^{3,4,6,8–10,13,14,16,19,20,22,23,25–28,31,37–39} were prospective cohort trials (23 of 28 studies; 82%). There were three retrospective cohort studies. ^15,32,33 Of the 11 papers rejected from further consideration, reasons for rejection included the fact that the study did not cover evaluation of lactate measurement or that the study cohort was recruited in the emergency department not the ICU or that the underlying clinical diagnosis in the study cohort was too narrow and therefore results were not generalizable. In the main, studies used lactate measurements as a comparator test rather than as the reference test.

Analyte and matrix

Twenty six of 28 studies considered l-lactate in serum. One study used d-lactate measurement ³¹ and one study looked at luminal d- and l-lactate in colonic washings. ²⁸

Interpretation

Serum l-lactate measurement was found to be equal or superior to many other markers. However, there was a broad consensus among most studies that although high admission serum l-lactate concentration was associated with sequential organ failure and 28-day in-hospital mortality, there was too great a degree of overlap between concentrations in patients who survived and those who did not, such that a specific cut-off capable of predicting which patients were more likely to survive and which were more likely to die within 28 days with acceptable sensitivity and specificity could not be recommended.

One study applied ROCC analysis to rate of decline in d- and l-lactate concentrations as predictors of 28-day mortality and found that the area under the ROCC was 0.68 and 0.84 for d- or l-lactate concentrations, respectively. ³¹

The most common recommendation across all studies was to use normalization of serum l-lactate concentration as a target during goal-directed therapy because of the overlap of serum l-lactate concentrations between survivors and non-survivors.

Conclusions and recommendations

We found a lack of good-quality studies of the use of admission lactate concentrations as a prognostic marker in the ITU setting. There were few RCTs, and the two RCT studies that we identified were not focused on lactate measurement.

The broad consensus of evidence from the studies that were reviewed was that there is an association between the highest serum l-lactate concentrations and eventual poorer outcome. Although all studies suggested that measurement of admission and serial serum l-lactate concentrations could be helpful in the ICU, no specific cut-off concentrations were presented which had strong evidence to support them and there was no consensus across studies.

It was concluded that there is insufficient good-quality evidence to suggest that measuring admission serum l-lactate concentrations is useful for the specific purpose of prognostication in ICU. The practice cannot be recommended as there is currently no explicit cut-off concentration for admission serum l-lactate concentration capable of accurately predicting clinical outcome.

However, there was stronger evidence that patients may benefit if goal-directed therapy includes a target to return serum l-lactate concentrations to the reference range. This review suggests that there is a need for well-designed RCTs of goal-directed therapy using serum l-lactate concentrations to guide interventions, to define specific cut-off concentrations and quantify expected rates of decline.

There were too few studies using d-lactate concentrations or measuring d- and/or l-lactate in atypical body fluids for us to make specific recommendations, but in those few studies, a possible role over and above that identified for serum l-lactate concentrations was suggested. This review concludes that the role of d-lactate and use of non-standard biological fluids warrants further investigation.

DECLARATIONS

Competing interests: None.

Funding: None.

Ethical approval: Not required.

Guarantor: CC.

Contributorship: All authors made an equal contribution to the review. Each attended a half-day EBLM workshop in May 2010 sponsored by the Association for Clinical Biochemistry. All authors attended a further whole day at Royal Manchester Children's Hospital in September 2010, February 2011 and May 2011, at which the protocol was drafted, the search strategy was devised and performed, and the initial review of all abstracts was undertaken. Each author obtained and formally appraised one quarter of the publications identified using the methodology as described.

Acknowledgements: The following individuals and groups provided practical help and advice for which the authors wish to express their gratitude: Professor Christopher Price, Dr Mark Sleeman, Sally Benton, ACB Scientific Committee and ACB Education Committee.