Effect of Timely Lactate Measurement on In-hospital Mortality among Adults with Hypotension and Hyperlactatemia: An Observational Study on Two Cohorts

Background: Whether patients presented with hypotension and hyperlactatemia can benet from timely lactate measurement and further lactate-guide resuscitation were not fully understood. Methods: This was a retrospective observational study based on the data from the Medical Information Mart for Intensive Care (MIMIC)-III Database and the eICU Collaborative Research Database (eICU). Patients with hypotension (dened as a minimal systolic blood pressure ≤ 90 mm Hg or minimal mean arterial pressure ≤ 65 mm Hg or requiring any vasopressors support during the rst 24 h after ICU admission) and hyperlactatemia (dened as an initial lactate level > 2.0 mmol/L after ICU admission) were eligible.The primary exposure was the timely lactate measurement, which was dened as an initial lactate level measured within 1 h after ICU admission. The primary outcome was in-hospital mortality. The statistical approaches included multivariate regression, propensity score matching (PSM) and an inverse probability of treatment weighing (IPTW) and causal mediation analysis (CMA) were utilized to elucidate the relationship between timely lactate measurement and in-hospital mortality. Results: A total of 9978 patients were identied, of which 4257 in MIMIC-III and 5721 in eICU. Timely lactate measurement was associated with lower risk-adjusted in-hospital mortality both in MIMIC (OR 0.70 (95%CI 0.58-0.85; p<0.001)) and eICU (OR 0.75 (95%CI 0.64-0.88; p<0.001)). Time to initial intravenous uid (IVF) in MIMIC mediated 6.7% (95%CI 1.4%-38%; p<0.001) of the benecial effect of timely lactate measurement (p<0.001 for average causal mediation effect (ACME)) in terms of in-hospital mortality. Finally, delayed initial lactate measurements are also associated an increased in-hospital mortality in MIMIC and eICU. Conclusions: Timely lactate measurement is associated with a lower risk-adjusted in-hospital mortality among patients with hyperlactatemia, which was proportional mediated through shortening the time to IVF. Delay in initial lactate measurement showed a positive association with in-hospital mortality.

Introduction Since lactate measurement was rst described by Scherer in 1843, blood lactate level measurement was performed widely in critically ill patients, especially in patients with shock [1,2]. On the one hand, elevated lactate levels in shock is a predictor of poor clinical outcome as a biomarker of tissue hypoperfusion [3], On the other hand, lactate level is an effective parameter for guiding resuscitation during the early phase of shock [4][5][6][7]. Furthermore, clinical therapy according to the elevated lactate level is believed to be associated with increased risk of death when delayed [8]. Thus, shorter the time to initial lactate level measurement is of vital importance to the management of shock. However, considerable controversies still exist in regarding how rapidly measurement of lactate level should be performed in critically ill patients with shock.
Few studies have focused on assessing optimal time to initial lactate measurement after intensive care unit (ICU) admission and its effect on outcome in septic patients. Pruinelli et al. [9]concluded that for patients with severe sepsis or septic shock, a delay of 20 minute in initial lactate level measurement signi cantly increased the risk of in-hospital mortality, and the lactate recommendation had minimum statistically signi cant time compared with other 3-hour bundle recommendations. The latest sepsis campaign guideline [10]also emphasized that the initial lactate level should be performed within 1 h despite of no strong evidence of bene t. Notably, even though many groups have demonstrated in circulatory shock that hyperlactemia was associated with elevated morbidity and mortality [11][12][13][14][15], but whether all patients presented with circulatory shock or unstable hemodynamics can bene t from timely lactate level measurement and further lactate-guide resuscitation were not fully understood.
The present study was undertaken to elucidate the relationship between timely lactate measurement (within 1 h after ICU admission) and outcomes and in uences of timely lactate measurement on interventions in patients with hypotension and hyperlactatemia (> 2.0 mmol/L). In addition, we attempted to characterize the association of delays in initial lactate measurement with mortality to further understand the effectiveness of timely lactate measurement.

Study design
We conducted a retrospective cohort study based on two databases from critically ill patients admitted to the ICU, the Medical Information Mart for Intensive Care (MIMIC)-III [16], and the eICU Collaborative Research Database (eICU) [17].

Study population
The patients in the MIMIC-III and eICU databases who were aged 18 years or older were eligible if they had (1) hypotension (de ned as a minimal systolic blood pressure ≤90 mm Hg or minimal mean arterial pressure ≤65 mm Hg or requiring any vasopressors support during the rst 24 h after ICU admission) and (2) hyperlactatemia (de ned as an initial lactate level > 2.0 mmol/L after ICU admission). Based on the setting of previous researches, we excluded patients who spent less than 48 h in the ICU and patients without any measurement of lactate level within 24 h after ICU admission. Additionally, we only analyzed the rst ICU stay for patients who were admitted to the ICU more than once. The primary exposure was the timely lactate measurement, which was de ned as an initial lactate level measured within 1 h after ICU admission. The enrolled patients were further divided into two groups as follows: the initial lactate level was measured within 1 h after ICU admission (Lac-by1 group); the initial lactate level was measured more than 1 h after ICU admission (Lac-after1 group).

Data extraction
Time to initial lactate level measurements were calculated depend on the length of admission to ICU. Other variables of interest were also extracted from the MIMIC and eICU using structured query language (SQL) (Additional le 1: Additional Methods).

Outcomes
The primary outcome of our study was in-hospital mortality. Secondary outcomes including the time to initial vasopressor administration (hours); time to initial intravenous uid (IVF) treatment (hours); volume Timely lactate measurement was used as a binary variable for the primary analysis. Multivariate regression was selected to characterize the relationship between timely lactate measurement and inhospital mortality. In the eICU, to responsible for potential changes of clinical practice in different centers, an analysis including center as a random factor in mixed-effect models was performed. To avoid bias induced by missing data, the analysis of the primary outcome was duplicated after multiple imputation. Propensity score matching (PSM) [18]and propensity score-based inverse probability of treatment weighing (IPTW) [19]were also used for adjusting the covariates to ensure the robustness of our ndings.
To explore whether the effect of timely lactate measurement on the primary outcome is mediated by the interventions, we used causal mediation analysis (CMA) [20]to searching for key interventions including the time to IVF and time to initial vasopressor administration that could be linked to positive outcomes after timely lactate measurement (Additional le 1: Additional Methods).
Time to initial lactate measurement was then used as a continuous variable for the other analysis.
Restricted Cubic Spline were used to detect the possible nonlinear relationship between the risk of inhospital death and time to initial lactate measurement. Moreover, we analyzed the time to initial lactate measurement partitioned into 5 quantiles to clarify the relationship further in two cohorts.
All statistical analyses were performed using the RStudio (version 1.2.5019), and p < 0.05 was considered statistically signi cant.

Baseline characteristics
After reviewing 53423 MIMIC-III adult admissions, a total of 4257 unique patients admitted from 2001 till 2012 both ful lled the de nition of hypotension and hyperlactatemia. In eICU, we had enrolled 5721 unique patients from 200859 admissions admitted from 2014 till 2015 in our study. The ow diagrams of study patients are presented in Figure S1 and S2 (Additional le 1), respectively.
The baseline characteristics of two cohorts are summarized in Table 1 and Table S1 (Additional le 1); characteristics of ICUs involved are presented in Table S2 (Additional le 1). In MIMIC-III, the minimum mean arterial pressure (MAP) was 54 mmHg (IQR, 48-60 mmHg), and the initial lactate level was 3.

Time to initial lactate level measurements
Distributions of time to initial lactate level measurements in two cohorts are seen in Figure S3 Table S3 and S4 (Additional le 1).

Primary outcome
Results of the univariate analyses in two cohorts are presented in Table S5 (Additional le 1), the full multivariate logistic regression analyses are showed in Figure 1 and Table S6 (Additional le 1), the adjusted odds ratio (OR) was 0.70 (95% CI 0.58-0.85; p<0.001) in MIMIC and 0.75 (95%CI 0.64-0.88; p<0.001) in eICU. Moreover, this bene cial effect was not only con rmed after multiple imputation (Additional le 1: Table S7 and S8), but also in the models when using PSM and IPTW (Additional le 1: Table S3 and S4, Figure S4 and S5) for adjusting the covariates in two cohorts. On average, the completion of initial lactate measurement within 1 h after ICU admission was associated with lower riskadjusted in-hospital mortality for patients with hypotension and hyperlactatemia both in MIMIC and eICU.
Subgroups analysis were completed in both cohorts (Additional le 1: Figure S6). Brie y, timely lactate measurement was associated with improved outcome despite of the different initial lactate level (2-4; ≥4), the Sequential Organ Failure Assessment (SOFA) score (<10; ≥10) and minimal MAP (<65; ≥65), while no statistically signi cant interaction were detected. Secondary outcomes with propensity score matched cohorts Therapeutic interventions including vasopressor and IVF that might mediate the bene cial effect of timely lactate measurement on in-hospital mortality were investigated. First, patients in the Lac-by1 group had a shorter time to initial any vasopressor administration in MIMIC and eICU, and a shorter time to initial norepinephrine in MIMIC when administered. Second, compared with Lac-after1 group, Lac-by1 group had a signi cant shorter time to IVF and received more IVF within 6 h in two cohorts ( Table 2).

Test of mediation
We further con rmed whether the bene cial effect of timely lactate measurement on in-hospital mortality is mediated through therapeutic interventions. After causal mediation analysis, we found that the time to IVF in MIMIC mediated 6.7% (95%CI 1.4%-38%; p<0.001) of the bene cial effect of timely lactate measurement (p<0.001 for average causal mediation effect (ACME)) ( Figure 2). However, when using time to IVF in eICU or time to initial vasopressor administration in MIMIC and eICU as mediator, the mediation effect was insigni cant (Additional le 1: Table S9).Thus, survival bene t of the timely lactate measurement was proportional mediated through shortening the time to IVF.
Delayed initial lactate measurement and in-hospital mortality Figure 3 shows the increase in the adjusted odds of in-hospital mortality as a contribution of progressive percentiles of time to initial lactate measurement in MIMIC ( Figure 3A) and eICU ( Figure 3B).

Discussion
Our investigation yielded three major ndings: (1) Timely lactate measurement (within 1 h after ICU admission) was associated with lower risk-adjusted in-hospital mortality in patients with hypotension and hyperlactatemia; (2) Survival bene t of the timely lactate measurement was proportional mediated through shortening the time to IVF; (3) Delayed initial lactate measurement showed a positive association with in-hospital mortality.
This is the rst investigation to proof the bene t of timely lactate measurement in all patients with unstable hemodynamics (hypotension and hyperlactatemia) rather than just for septic. In our study, despite patients who had initial lactate measurements completed within 1 h had higher SOFA scores, higher initial lactate level and receive more mechanical ventilation and vasopressors, the in-hospital mortality was signi cantly lower among patients who had initial lactate measurements completed within 1 h after adjustment for confounding. This parallels the ndings of previous limited researches, speci cally for septic patients. A retrospective study conducted by HC et al. [21] declared that early lactate measurement (within 1 h after ICU admission) was associated with lower odds of 28-day mortality among septic patients with elevated serum lactate level. Similarly in Han et al.'s study [22], septic patients with delayed lactate measurements (lactate samples drawn more than 3 h after meeting severe sepsis criteria) demonstrated the highest in-hospital mortality than did patients who had lactate samples drawn within 3 h. On the whole, our results accentuated an essential role of timely lactate measurement for all patients with unstable hemodynamics admitted to ICU.
Interestingly, the de nition of "delay" was depend on the length of exposure to severe sepsis in Han's study, while the actual onset time of severe sepsis is unknown truthfully, which might occurred outside the ICU, moreover, the etiologies for patients with unstable hemodynamics were not always explicit [23,24]. Although it's possible that patients would have received timely resuscitation regardless of the etiologies and whether the lactate level was measured or not, identifying patients could delay the lactate measurement and interventions. Thus, the "timely" in present study was beginning with the onset of ICU admission to led clinicians consider lactate draw earlier for all patients with suspected unstable hemodynamics. Additionally, Han et al. [22] detected a signi cant relationship between delay in initial lactate measurement and in-hospital death only in patients with an initial lactate value > 2.0 mmol/L, raising the question that the initial lactate level or other clinical parameters associated with the severity of diseases might cause interaction effects between timely lactate measurement and mortality. However, no signi cant interactions were detected for initial lactate level, SOFA score and minimal MAP in both cohorts. Actually, there was no highly recognized factors that can represent the disease severity, and the mechanism underlying the interaction between these factors and timely lactate measurement is little known. More research is required to develop a deeper understanding of the interaction effects above.
What's more attractive is to investigate which factors were contribute to the bene t effects of timely lactate measurement in terms of in-hospital mortality. Therapeutic interventions have been shown to mediate this bene t effect in previous studies, including uid resuscitation, administration of vasoactive agents, both of which were the important components of resuscitation in circulatory shock [25]. According to a randomized controlled trial by Jasen et al. [26], patients with lactate levels > 3.0 mmol/L in lactate group (with the objective of decreasing their lactate levels by 20% or more per 2 h in the initial 8 h of their ICU stays) received signi cantly more uids and more vasodilator therapy during 0-8h and had a lower hospital mortality compared to the control group. Besides, the time to initial therapeutic interventions were equally important. HC et al. [21] demonstrated that early lactate measurement (within 1 h after ICU admission) was associated with early antibiotic therapy and vasopressors administration in septic patients with lactate levels > 2.0 mmol/L, and further con rmed that the time to initial vasopressors was proportional mediated the survival bene t effect of early lactate measurement. On the contrary, in our study, although patients in Lac-by1 group had a shorter time to IVF and vasopressors in both cohorts, we only found a signi cant mediation effect of time to IVF in MIMIC using CMA.
Adequate and timely uid therapy is an essential part of resuscitation of circulatory shock even in cardiogenic shock. A Retrospective cohort study by Kuttab et al. [27] found that septic patients who were failed to achieving a 30mL/kg crystalloid bolus within 3 hours of sepsis onset was associated with increased odds of in-hospital mortality, irrespective of patients who were elderly, end-stage renal disease, heart failure or documented volume "overload". Still in Han et al.'s study [22], patients with lactate levels > 2.0 mmol/L in draw lactate group had a shorter time to IVF and a lower mortality compared with patients in delayed lactates group. Even though we detect no signi cant mediation effect of time to IVF in eICU, we should not underestimate the value of timely IVF in the management of shock, since our study is a retrospectively study performed on electronic health record data and require further investigate. Studies undertaken so far provide con icting evidence concerning the impact of time to initial vasopressor on mortality in circulatory shock [28][29][30], the effectiveness of vasopressor depends on the type and severity of shock, and most patients would have vasopressors administrated prior to ICU admission in two cohorts. These might account for the insigni cant mediation effect of time to initial vasopressors administration.
The strength of study was the use of MIMIC-III and eICU databases which contain comprehensive and high-quality data of well-de ned and characterized ICU patients in 147 different hospitals from the USA. We also used model of PSM and IPTW to minimize confounding factors and robust our results. It is di cult to elucidate clearly on causality in a retrospective analysis, we apply the CMA to cover this shortage. The present research is also limited by several factors as follows: First, the diagnosis of circulatory shock is based on clinical, hemodynamic and biochemical markers, which summarized into three parts: systemic hypotension, clinical signs of tissue hypoperfusion and hyperlactatemia [25], it's impossible to assess the clinical signs of tissue hypoperfusion in our study, hence we only use hypotension and hyperlactatemia to represent the patients with circulatory shock; Second, our study ranging from 2001 till 2015, the versions of the bundles might have changed during the period, and the results may not be adapted to current practice, however, our results were adjusted for the admission period, in MIMIC, since we couldn't get the exact year of patients' admission, we divided patients into two groups in terms of admission year (before 2008 and 2008-2012).Third, lactate guiding resuscitation are intricacy in clinical ,more than just uid and vasoactive agents. The etiologies of shock, the severity of shock and unmeasured confounding factors such the reaction and the decision of the clinicians on duty are all involved in the management of shock and need to be explored in future. Finally, lactate is an imperfect marker of anaerobic metabolism [31], and lactate cannot be used to guide resuscitation alone, tissue hypoxia should be assessed in a combined analysis including ScvO 2 or indices derived from the veno-arterial carbon dioxide pressure gradient.

Conclusion
In our study, we found that patients with hypotension and hyperlactatemia in Lac-by1 group demonstrated a lower risk-adjusted in-hospital mortality, and this survival bene t was proportional mediated through shortening the time to IVF. Timely lactate measurement may hence be useful in all patients with circulatory shock or unstable hemodynamics. Association between timely lactate measurement and in-hospital mortality. The odds ratios and 95% con dence intervals (error bars) in both cohorts were calculated dependent on method of covariate adjustment. CBPS: covariate balancing propensity score, IPTW: inverse probability of treatment weight. Causal mediation analysis for time to IVF. The solid line represents the Lac-by1, and the dashed line represents the Lac-after1.