Septic shock is an important admission diagnosis among intensive care patients and is classically defined by the presence of refractory hypotension (1). It is also a major cause of mortality (2). Thus, identifying those patients at greatest risk of death is essential in order to achieve better outcomes.
The original early goal-directed therapy (EGDT) study (3) promoted the use of blood lactate concentrations ≥ 4 mmol/L in the absence of refractory hypotension to identify patients with “cryptic shock.” The Surviving Sepsis Guidelines (4) later adopted hyperlactatemia as a trigger for initiation of EGDT. However, in septic patients, limited data exists to support the premise that isolated hyperlactatemia reflects a pathophysiological state equivalent to refractory hypotension in terms of patient outcomes.
Hyperlactatemia appears to reflect the imbalance of complex biological events such as global or regional hypoperfusion, increased glycolysis, alterations in pyruvate dehydrogenase activity, the preferential generation of lactate as an energy source, and decreased lactate clearance (5). A post-hoc analysis (6) of a small cohort of septic shock patients suggested that patients with diabetes, end-stage renal failure, and intra-abdominal sepsis were more likely to present with cryptic shock and that patients with cryptic shock had similar mortality to those with overt septic shock. Another observational study questioned whether hypotension without hyperlactatemia is true septic shock since lower mortality was observed in patients without hyperlactatemia (>2.4 mmol/L) (7).
Thus, it remains unclear if isolated hyperlactatemia patients have different underlying diagnoses, receive different care or experience different outcomes compared with those with isolated hypotension. Importantly, it is unclear if those with isolated hyperlactatemia require a different resuscitation strategy compared with those with refractory hypotension.
The aim of this study was to compare the baseline characteristics, physiological parameters, treatment features, 90-day mortality, and other clinically relevant outcomes among Australasian Resuscitation in Sepsis Evaluation (ARISE) (8) trial participants enrolled on the basis of isolated hyperlactatemia compared with isolated refractory hypotension. Specifically, we hypothesized that participants with isolated hyperlactatemia would receive similar treatments and have similar outcomes to those of participants with refractory hypotension.
PATIENTS AND METHODS
This study was a secondary analysis of the Australian and New Zealand Intensive Care Society Clinical Trials Group and Australasian College for Emergency Medicine-endorsed ARISE trial, a multicenter randomized study comparing EGDT to usual care in patients presenting to the emergency department (ED) with early septic shock. The ARISE study enrolled 1,600 adult participants from 51 sites, the majority of which were Australasian with the remainder in Finland, Hong Kong, and Ireland. The trial protocol, statistical analysis plan, and main study results have been published previously (8).
Patients were eligible for inclusion in ARISE if they were 18 years or over, had confirmed or suspected infection with two or more systemic inflammatory response criteria and refractory hypotension, hyperlactatemia, or both physiological states. Refractory hypotension was defined as a systolic blood pressure of less than 90 mm Hg or a mean arterial pressure of less than 65 mm Hg after an intravenous fluid challenge of 1,000 mL or more administered within a 60-min period. Hyperlactatemia was defined as a blood lactate level of 4.0 mmol/L or more (8).
Patients were eligible for inclusion in this secondary study only if either isolated hyperlactatemia or isolated refractory hypotension was present as the trigger for randomization. We excluded patients with combined refractory hypotension and hyperlactatemia as the trigger for randomization as their combined presence made separate assessment of each individual trigger impossible.
We applied the following endpoint definitions to our analysis:
- 90-day mortality: all-cause mortality within 90 days of randomization;
- ICU mortality: all-cause mortality during first ICU admission;
- in-hospital mortality: all-cause mortality during first hospital admission;
- total intravenous fluid administration: the combined volume of any crystalloid or colloid administered in each observation period (0–6, 7–24, and 25–72 post-randomization);
- hospital and intensive care (ICU) (if admitted) length of stay (LOS): the duration of the first admission;
- duration of mechanical ventilation (invasive and/or noninvasive): the post-randomization duration of the first episode of mechanical ventilation to cessation of the last episode, including intervening periods of unsupported ventilation; and
- duration of vasopressor administration: the post-randomization duration from initiation of the first infusion to cessation of the last infusion, including periods without any vasopressor support.
We summarized continuous and discrete variables by mean and standard deviation (SD) or medians and interquartile range (IQR) as appropriate, and binary variables as proportions. We performed statistical significance testing of baseline characteristics using Student t test, Wilcoxon rank sum, Fisher exact test, or chi-squared tests as appropriate.
A propensity score model was constructed to estimate the probability of developing isolated hyperlactatemia or isolated refractory hypotension. Inverse probability of treatment weighting was used to create a “pseudo-population” in which the baseline covariates were balanced (9, 10). The same weights were used for all analyses. All variables with complete data were included in the propensity model. To avoid a large reduction in sample size, further variables were included if less than 10% of data were missing and they were associated with shock type at a P value of <0.1. The propensity score approach involved the development of a logistic regression model to predict shock type, either isolated refractory hypotension or isolated hyperlactatemia, using the following baseline variables: age, gender, weight, usual residence, Charlson comorbidity index (CCI) score, APACHE II score, mechanical ventilation at baseline, presence of chronic medical conditions, concurrent medical conditions, albumin and hemoglobin concentrations, Glasgow Coma Scale, site of infection, hospital location and hospital type. No attempt was made to balance those baseline measurements that were likely to be a consequence of hypotension or hyperlactatemia, including time to randomization, bicarbonate concentration, vasopressor use, volume of IV fluid administered, respiratory rate, heart rate, or blood pressure. In brief, we assessed imbalance between those with isolated refractory hypotension and those with isolated hyperlactatemia using standardized differences (9, 10). An iterative process was used to achieve <10% imbalance for all variables (Table 1 and Supplemental Digital Content 1, https://links.lww.com/SHK/A573). The “common support” condition was imposed, meaning that observations were removed if there were no participants of comparable propensity score in the other group.
Further details and the final propensity score model are given in the Supplemental Digital Content 2, https://links.lww.com/SHK/A573. The proportion of the variables with missing data are summarized in Supplemental Digital Content 3, https://links.lww.com/SHK/A573.
Associations between shock and 90-day mortality; first admission ICU mortality; first admission in-hospital mortality; ICU admission; any use of ventilation; and any use of vasopressors were modeled using a generalized linear model for the binomial distribution with log link function to directly estimate relative risks (RR). ICU and hospital length of stay, vasopressor use, and mechanical ventilation were modeled using competing risks analysis, death being the competing event. Competing risks analysis takes into account that death and clinical improvement are mutually exclusive and therefore allows the inclusion of participants who died in the analysis (11). We estimated adjusted and unadjusted cumulative incidence plots for each group. We presented results as comparisons between groups using sub-hazard ratios (SHR) with 95% confidence intervals (CI). The sub-hazard is the instantaneous risk of discharge given that the patient remains alive. We tested the proportional hazards assumption by including shock as a time varying covariate in the competing risks models.
Sensitivity analyses were performed comparing ICU-free survival, hospital-free survival, ventilator-free survival, and vasopressor-free survival (up to 90 days) in the two groups using linear regression and inverse probability of treatment weighting.
We analyzed intravenous fluid administration for each study observation period using a generalized linear model with log link function and binomial distribution family. The results are presented as risk ratios (RR) with 95% CI. For those who received fluid, fluid resuscitation volume was log transformed and modeled using linear regression. We presented results as ratios of geometric means with 95% CI. We performed all calculations with Stata Version 14 (StataCorp, College Station, Tex).
Participants and baseline characteristics
Overall, 1,332 participants were included in this analysis based on the presence of either isolated hyperlactatemia (478 [35.9%]) or isolated refractory hypotension (854 [64.1%]) (Fig. 1). We excluded an additional 258 participants with combined hyperlactatemia and refractory hypotension according to study protocol. Data for propensity-score estimation were missing for 409 participants (31%); 23 participants (1.7%) had propensity scores outside the region of common support and 3 participants were missing 90-day mortality data, leaving a final cohort of 897 participants (330 isolated hyperlactatemia and 567 isolated refractory hypotension) for the propensity analysis.
The baseline characteristics for participants meeting either the isolated hyperlactatemia or isolated refractory hypotension criterion, but not both, are shown in Table 1 and Supplemental Digital Content 1, https://links.lww.com/SHK/A573.
Participants with isolated hyperlactatemia and isolated refractory hypotension were of similar age and gender. However, isolated hyperlactatemia participants had a higher chronic health burden (liver disease, diabetes) and overall had more concurrent medical conditions (P = 0.02) at enrolment (Supplemental Digital Content 1, https://links.lww.com/SHK/A573). They also had a higher APACHE II score (16.2 [6.4] vs. 14.5 [6.4] for isolated refractory hypotension), higher hemoglobin and albumin concentrations and a lower serum bicarbonate concentration (P < 0.001 for all). There were also some geographic differences in the distribution of isolated hyperlactatemia with Hong Kong sites having a lower proportion of isolated hyperlactatemia participants than other sites (P = 0.02) (Supplemental Digital Content 1, https://links.lww.com/SHK/A573).
Infection site was similar across the two groups (Table 1), except for a higher rate of central nervous system infections in participants with isolated hyperlactatemia (12 [2.5%] vs. 5 [0.6%], P = 0.003). There were no differences with respect to the distribution of causative organisms (Supplemental Digital Content Digital Content 4, https://links.lww.com/SHK/A573). However, the volume of intravenous fluid administered prior to enrolment was less for participants with isolated hyperlactatemia (18.8 [10.4, 30.6] mL/kg vs. 33.5 [23.5, 46.2] mL/kg; P < 0.001). Isolated hyperlactatemia participants were also more likely to receive mechanical ventilation prior to enrolment (16.7% vs. 10.1%; P < 0.001) but less likely to receive a vasopressor infusion (5.4% vs. 19.1%; P < 0.001). The time from emergency department presentation to randomization was shorter for participants with isolated hyperlactatemia (P < 0.001).
Standardized differences between all variables thought to be confounders were reduced to less than 10% after inverse propensity score weighting (Table 1 and Supplemental Digital Content 1, https://links.lww.com/SHK/A573).
The unadjusted risk of 90-day mortality in isolated hyperlactatemia participants (23.1%; 110 of 478 participants) was almost twice that seen in the isolated refractory hypotension group (12.3%; 105 of 854 participants) (Table 2). After propensity score weighting, the RR of mortality was 1.7 (95% CI 1.2, 2.5, P = 0.003).
The unadjusted mortality rate for the first ICU admission was higher in the isolated hyperlactatemia participants (16.7%; 45 of 270 participants) compared with isolated hypotension participants (6.2%; 46 of 747) with a propensity weighted RR of 2.3 (95% CI 1.4, 3.7; P = 0.001).
Unadjusted mortality during the first hospital admission was also higher in the isolated hyperlactatemia participants (17.2%; 57 of 331) compared with isolated hypotension participants (9.1%; 77 of 847) with a propensity weighted RR of 1.7 (95% CI 1.1, 2.7; P = 0.01).
Isolated hyperlactatemia participants were less likely to be admitted to ICU (unadjusted RR 0.95 95% CI 0.90, 0.99, P = 0.02) (Table 2). However, the propensity-weighted cumulative incidence plot found that isolated hyperlactatemia participants had a longer ICU length of stay, after accounting for their higher ICU mortality (Fig. 2). The probability of being discharged from ICU alive was also lower for participants with isolated hyperlactatemia (propensity weighted SHR 0.77 [95% CI 0.64, 0.92; P = 0.005]) (Table 3).
Hospital length of stay, after accounting for in-hospital mortality, was longer among isolated hyperlactatemia participants (Fig. 2). Such participants were also less likely to be discharged alive from hospital (propensity weighted SHR 0.79 [95% CI 0.66, 0.95; P = 0.01] respectively) (Table 3).
Mechanical ventilation was more commonly employed in participants with isolated hyperlactatemia (48.7% vs. 33.3%; P < 0.001) (Table 3), although the risk of ventilation was similar after propensity weighting (adjusted RR 1.20 [95% CI 0.97, 1.46; P = 0.09]). There was no difference in the duration of mechanical ventilation, the propensity weighted SHR for cessation of mechanical ventilation for participants with isolated hyperlactatemia was 0.75 (95% CI 0.55, 1.01, P = 0.06) (Table 3).
In contrast, isolated hyperlactatemia participants were less likely to receive vasopressors (58.8% vs. 74.0%; P < 0.001) (adjusted RR 0.76 [95% CI 0.67, 0.85; P < 0.001]) but duration of vasopressor support was longer (Table 2, Fig. 2). The propensity-weighted SHR for cessation of vasopressor infusion for participants with isolated hyperlactatemia was 0.63 (95% CI 0.50, 0.79; P < 0.001) (Table 3).
Isolated hyperlactatemia patients had shorter mean ICU-free, hospital-free, ventilator-free, and vasopressor-free survival (Supplemental Digital Content 5, https://links.lww.com/SHK/A573). This was consistent with the analysis based on competing risks regression.
Participants with isolated hyperlactatemia were less likely to receive fluid from 25 to 72 hours post-randomization than the isolated refractory hypotension group on unadjusted analysis (Supplemental Digital Content 5, https://links.lww.com/SHK/A573). However, after propensity-weighted linear regression, the isolated hyperlactatemia group received a greater volume of fluid in the first 6 h compared with the isolated refractory hypotension group (Supplemental Digital Content 6, https://links.lww.com/SHK/A573) but less fluid from 25 to 72 h post-randomization.
We performed a detailed secondary analysis of the ARISE trial. We found that participants enrolled on the basis of isolated hyperlactatemia had a higher 90-day mortality; were less likely to be discharged alive from both ICU and hospital; had longer ICU and hospital duration of stay; received more intravenous fluid in the first 6 h; and received mechanical ventilation more often than participants with isolated refractory hypotension. Moreover, although vasopressors therapy was less common, duration was longer. Finally, these outcome differences persisted after propensity score adjustment for key baseline covariates, including severity of illness, management prior to randomization and chronic health conditions.
Relation to previous findings
Hyperlactatemia has been shown to predict worse outcomes in sepsis (12, 13). However, few studies have specifically compared patients with either isolated hyperlactatemia (“cryptic shock”) or isolated refractory hypotension. In a post-hoc analysis of 300 patients from a randomized trial of severe sepsis, mortality was similar for cryptic shock (lactate ≥4 mmol/L and normotension) and refractory hypotension (6). However, there were only 53 patients in the cryptic shock group. In contrast, a recent cohort study found that patients with a mean lactate >2 mmol had worse ICU survival than those with lactate <2 mmol, regardless of whether their mean arterial pressure was less than or greater than 65 mm Hg (14).
In a study of 19,945 patients from the Surviving Sepsis Campaign Database, the combination of a lactate ≥4 mmol/L and hypotension had higher odds of in-hospital mortality compared with isolated hyperlactatemia or hypotension (13). After adjusting for confounders, however, no difference in mortality was found between isolated hyperlactatemia and isolated hypotension. Moreover, the unadjusted mortality rate in hypotensive patients with a blood lactate <4 mmol/L (isolated hypotension) was 29.3%, compared with 12.3% in our patients, suggesting that populations or their treatment or both were markedly different. Moreover, many such lactate measurements occurred more than 6 h following presentation, whereas our study categorized participants according to baseline lactate measurements obtained at a mean of 1.3 h after ED presentation. In a subsequent analysis of 18,840 patients in the Surviving Sepsis Campaign Database, patients with isolated hyperlactatemia (>4 mmol/L) in the first 24 h had a similar unadjusted mortality (29.9% [555/1,856]; [95% CI 27.8, 32.0]) to those with isolated hypotension (30.1% [1,198/3,985]; [95% CI 28.6, 32.5]) (15). However, on regression analysis, those with isolated mild hyperlactatemia (>2 mmol/L) had higher odds of mortality than those with isolated hypotension. Unfortunately, no direct comparison was reported for patients with an isolated lactate greater than 4 mmol/L and those with isolated hypotension. Moreover, 790 patients with a lactate >4 mmol/L were excluded from analysis because they did not receive fluids as per the Surviving Sepsis Guidelines, thus creating a significant selection bias effect. Nonetheless, after adjustments for available variables, there was an almost linear increase in the odds ratio for mortality with increasing blood lactate concentrations.
There are reasons why hypotension alone may be an inadequate marker of a septic shock state (16). Patients with chronic hypertension may require higher systemic blood pressure to maintain normal perfusion (17) and therefore may have hypoperfusion at normal blood pressures. Other patients may maintain adequate tissue perfusion at nominally low blood pressure or their hypotension may be exacerabted by sedation and therefore not accurately reflect the severity of sepsis. In contrast, normotensive hyperlactatemic patients appear to have greater disorders of coagulation and of microcirculation (7).
Implications of study findings
The findings of our study imply that an isolated blood lactate ≥4 mmol/L identifies patients at greater risk of death than refractory hypotension alone and that a lactate level >4 mmol/L should be considered an important component of any defintion of septic shock even in the absence of refractory hypotension. Moreover, they imply that measurement of lactate in patients with suspected or proven infection presenting to the Emergency Department is a key risk stratification tool, identifying a distinct subgroup of severe sepsis patients that may require different treatment strategies.
Strengths and limitations
To our knowledge, this is the largest study to directly compare isolated hyperlactatemia to isolated refractory hypotension in patients with severe sepsis. The main strength of this study is that it is based on a large prospectively collected dataset from a well-designed and executed mutlticentre trial with independent data monitoring and prespecified inclusion criteria. We identified factors that predispose a patient to develop either isolated hyperlactatemia or hypotension and used propensity score methods to balance these factors. Propensity score methods may yield more precise estimates of the effect of the exposure on the outcomes and be more robust to model misspecification than conventional logistic regression models when the number of outcomes is small relative to the number of confounders (18). Another strength was the use of competing risk analysis to account for death in time to event estimation. Alternatives, such as excluding patients who died from the analysis or using convential survival techniques, would be unsatisfactory—the former would have introduced immortal time bias and the latter would have resulted in under-estimates of length of stay or treatment duration.
Our study carries some limitations. It is post-hoc and observational and addresses questions the original trial was not designed to answer. It only involved patients from the emergency department, even though sepsis in ward patients is also common and important (19). It cannot provide information on the cause of death. However, such determination is problematic in ICU patients (20). Although propensity score methods are able to force a degree of balance on measured covariates, we cannot exclude residual confounding by unmeasured covariates. In the ARISE standard care group, hyperlactatemic patients were twice as likely as those with refractory hypotension to be treated on the ward. Less intensive care for some hyperlactatemic patients may be a partial explanation for their worse outcomes. Information was also not collected about commonly used medications, which may be important (21).
Finally, by estimating the propensity scores using complete cases only, we excluded 409 subjects. This should not cause bias if the data are missing at random conditional on covariates in the propensity score model. It is not possible to test whether data are missing at random but the inclusion of a large number of covariates in the propensity score model makes this assumption plausible.
ARISE trial partcipants randomized on the basis of isolated hyperlactatemia had worse 90-day mortality, were less likley to be discharged from ICU or hospital alive; had longer ICU and hospital stay; more frequent use of mechanical ventilation and received vasopressors for longer than those with isolated refractory hypotension. These observations imply that a blood lactate ≥4 mmol/L is likely an important independent marker of disease severity in sepsis; that it can be used for risk stratification and that it may be a candidate for the defintion of septic shock, even when present in isolation.
The authors gratefully acknowledge the professionalism and commitment of all the nurses and clinicans who participated in the care of the ARISE trial partcipants.
1. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, Bellomo R, Bernard GR, Chiche JD, Coopersmith CM, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA
2016; 315 8:801–810.
2. Kaukonen KM, Bailey M, Suzuki S, Pilcher D, Bellomo R. Mortality related to severe sepsis
and septic shock among critically ill patients in Australia and New Zealand, 2000–2012. JAMA
2014; 311 13:1308–1316.
3. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, Peterson E, Tomlanovich M, Early Goal-Directed Therapy Collaborative G. Early goal-directed therapy in the treatment of severe sepsis
and septic shock. N Engl J Med
2001; 345 19:1368–1377.
4. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, Sevransky JE, Sprung CL, Douglas IS, Jaeschke R, et al. Surviving sepsis campaign: international guidelines for management of severe sepsis
and septic shock: 2012. Crit Care Med
2013; 41 2:580–637.
5. Suetrong B, Walley KR. Lactic acidosis in sepsis: it's not all anaerobic: implications for diagnosis and management. Chest
2016; 149 1:252–261.
6. Puskarich MA, Trzeciak S, Shapiro NI, Heffner AC, Kline JA, Jones AE, Emergency Medicine Shock Research N. Outcomes of patients undergoing early sepsis resuscitation for cryptic shock compared with overt shock. Resuscitation
2011; 82 10:1289–1293.
7. Hernandez G, Bruhn A, Castro R, Pedreros C, Rovegno M, Kattan E, Veas E, Fuentealba A, Regueira T, Ruiz C, et al. Persistent sepsis-induced hypotension without hyperlactatemia: a distinct clinical and physiological profile within the spectrum of septic shock. Crit Care Res Pract
8. Investigators TA. Group tACT. Goal-directed resuscitation for patients with early septic shock. N Engl J Med
2014; 371 16:1496–1506.
9. Williamson EJ, Forbes A. Introduction to propensity scores. Respirology
2014; 19 5:625–635.
10. Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med
2015; 34 28:3661–3679.
11. Pintilie M. [An introduction to competing risks analysis]. Rev Esp Cardiol
2011; 64 7:599–605.
12. Tang Y, Choi J, Kim D, Tudtud-Hans L, Li J, Michel A, Baek H, Hurlow A, Wang C, Nguyen HB. Clinical predictors of adverse outcome in severe sepsis
patients with lactate 2-4 mM admitted to the hospital. QJM
2015; 108 4:279–287.
13. Casserly B, Phillips GS, Schorr C, Dellinger RP, Townsend SR, Osborn TM, Reinhart K, Selvakumar N, Levy MM. Lactate measurements in sepsis-induced tissue hypoperfusion: results from the Surviving Sepsis Campaign database. Crit Care Med
2015; 43 3:567–573.
14. Houwink AP, Rijkenberg S, Bosman RJ, van der Voort PH. The association between lactate, mean arterial pressure, central venous oxygen saturation and peripheral temperature and mortality in severe sepsis
: a retrospective cohort analysis. Crit Care
15. Shankar-Hari M, Phillips GS, Levy ML, Seymour CW, Liu VX, Deutschman CS, Angus DC, Rubenfeld GD, Singer M, Sepsis Definitions Task F. Developing a new definition and assessing new clinical criteria for septic shock: for the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA
2016; 315 8:775–787.
16. Hernandez G, Castro R, Romero C, de la Hoz C, Angulo D, Aranguiz I, Larrondo J, Bujes A, Bruhn A. Persistent sepsis-induced hypotension without hyperlactatemia: is it really septic shock? J Crit Care
2011; 26 4: 435 e439–414.
17. Bindra J, Pham P, Chuan A, Jaeger M, Aneman A. Is impaired cerebrovascular autoregulation associated with outcome in patients admitted to the ICU with early septic shock? Crit Care Resusc
2016; 18 2:95–101.
18. Cepeda MS, Boston R, Farrar JT, Strom BL. Comparison of logistic regression versus propensity score when the number of events is low and there are multiple confounders. Am J Epidemiol
2003; 158 3:280–287.
19. Ou L, Chen J, Burrell T, Flabouris A, Hillman K, Bellomo R, Parr M. Incidence and mortality of postoperative sepsis in New South Wales, Australia, 2002–2009. Crit Care Resusc
2016; 18 1:9–16.
20. Ridgeon E, Bellomo R, Myburgh J, Saxena M, Weatherall M, Jahan R, Arawwawala D, Bell S, Butt W, Camsooksai J, et al. Validation of a classification system for causes of death in critical care: an assessment of inter-rater reliability. Crit Care Resusc
2016; 18 1:50–54.
21. Wright BJ, Weingart SD. beta-Blockers and lactate in sepsis: same canary but different mine. Crit Care Med
2015; 43 12:2691–2692.