Sepsis, defined as a dysregulated host response to infection resulting in organ dysfunction, is associated with significant morbidity and mortality . In the emergency department (ED), it is important to identify the patient with early sepsis to ensure timely treatment and appropriate disposition . This is challenging, given the dynamic and heterogeneous nature of sepsis and the lack of reliable objective diagnostic criteria.
Near-infrared spectroscopy (NIRS) is a noninvasive technique for estimating tissue oxygen saturation (StO2), the percentage of oxygenated haemoglobin in skeletal muscle . This is most commonly measured at the thenar eminence of the hand. NIRS may identify patients with impaired tissue perfusion and aid clinical recognition of early sepsis . Sepsis results in alterations to the function of the microcirculation and abnormal StO2 (normal range 75–90%) may be an indicator of this. Although impaired StO2 correlates with other indices of oxygenation [6–8], the clinical utility in the ED remains uncertain. Studies have investigated StO2 for the diagnosis of sepsis , severity/mortality [11–13] and to predict the need for intensive care admission .
New-onset organ dysfunction is the hallmark of sepsis, and is central to the recently revised consensus definition . In practice, this is defined by an increase in the sequential organ failure assessment (SOFA) score of 2 or more points from baseline. In addition, a bedside assessment, the ‘quick sequential organ failure assessment’ (qSOFA) score, which allocates one point for each of systolic blood pressure (SBP) less than 100 mmHg, respiratory rate (RR) of at least 22/min and altered mentation, has been promoted to identify a higher risk of poor outcome.
The aim of the present study was to investigate whether StO2 measurement is useful for sepsis diagnosis and to predict clinical outcomes (mortality and prolonged ICU admission) among a cohort of ED patients admitted to hospital with infection.
Patients and methods
Design and setting
This was a multicentre, observational study of patients with suspected infection presenting to three Australian EDs during the following continuous, overlapping periods: November 2012 to November 2015; April 2013 to April 2016; and September 2015 to June 2016. The study was registered prospectively with the Australian and New Zealand Clinical Trials Registry (ACTRN12612001229864).
Inclusion criteria were based on the 2001 international consensus definition for sepsis in place at the time of commencement of recruitment . This required the presence of at least 2 systemic inflammatory response criteria (temperature >38 or <36°C; pulse rate 90 beats/min; respirations >20 breaths/min; white blood cell count >12×109/l or <4×109/l) in the setting of suspected or proven infection. Patients in whom the treatment plan included administration of intravenous antibiotics and admission to hospital were assessed for eligibility during rostered research personnel hours. Exclusion criteria were age less than 18 years, administration of intravenous antibiotics before presentation (i.e. interhospital transfers), expected mortality because of another cause (e.g. malignancy) within 90 days or pre-existing limitation of care order or other comorbidity precluding invasive organ support.
Ethics and consent
The South Metropolitan Health Service Human Research Ethics Committee approved the study. Written informed consent was obtained from participants or from next of kin. Because of the low-risk and observational nature of the research, and the need to obtain study measurements in a timely manner, approval was granted for enrolment under a waiver of consent to obtain initial data when necessary, with full written informed consent being obtained subsequently.
Study procedures and measurements
Data collection at baseline included demographics, vital signs, Charlson comorbidity score (CCS) , blood results including blood gas (venous or arterial), interventions performed and the SOFA score. Trained research nurses obtained StO2 measurements at the thenar eminence using the Inspectra Spotcheck 300 NIRS device (Hutchinson Technologies Inc., Hutchinson, Minnesota, USA) according to a defined protocol. A vascular occlusion test (VOT) was not used. StO2 readings were recorded separately from clinical data and not disclosed to the treating staff. A single StO2 reading was taken at enrolment (T0) and repeated 3 h later (T3). Patients were followed up to day 28, with SOFA scores being recorded daily until day 3 (72 h from admission) unless discharged from hospital earlier. Patients discharged directly from the ED after enrolment, had been transferred from ED to another hospital or self-discharged before completion of care in the study were excluded because of the inability of follow-up for the primary outcome.
Outcome measures and sample size calculation
The primary outcome was sepsis, as defined as a SOFA score increase of 2 or more points within 72 h from admission. This was adopted in the updated definitions published in 2016 . We therefore made a minor modification to emphasize an increase from the baseline SOFA score to harmonize with the contemporary definition. This change was registered on 8 November 2016, after completion of recruitment, but before un-blinding and analysis. Baseline SOFA elements requiring a laboratory result (bilirubin, platelet count, creatinine) were defined as the lowest pre morbid or convalescent result. Where this was unavailable, baseline was assumed to be zero for that SOFA domain. Septic shock was defined as a requirement for vasopressors and a serum lactate of at least 2 mmol/l, or persistent hypotension and elevated lactate in those participants in whom a determination was made to limit care to the ward level only. The secondary outcomes were 28-day in-hospital mortality and requirement for ICU admission more than 72 h duration. Outcomes were physician-adjudicated, without knowledge of the StO2 result.
Sample size was based on a clinically meaningful odds ratio (OR) of 1.6 for the primary outcome for a dichotomized abnormal StO2 reading of less than 75% , assuming an incidence rate of sepsis of 25% in the study population. A sample of 300 patients has 0.9 power to detect this with α = 0.05.
Mean ± SD values for StO2 were compared between the outcomes of interest using t-tests. The predictive utility of StO2 was determined by area under receiver operating characteristic curves (ROC AUC). Multivariate logistic regression, with covariates including age, RR, SBP, Glasgow coma score (GCS), serum lactate and Charlson score, was performed using a backward-stepwise method to provide adjusted estimates of the association between StO2 and outcomes. Analyses were carried out using the StO2 at each time point, the mean of the two values and the difference between the T0 and T3 readings . The primary analysis was on an intention-to-treat basis, with a preplanned sensitivity analysis ‘perprotocol’ excluding any cases whose subsequent clinical course/diagnosis was not considered to be sepsis. Analyses were carried out using Stata v13 (StataCorp, College Station, Texas, USA).
The study enrolled 342 participants, of whom 323 were included in the primary analysis (Fig. 1). Of these, 143 (44%) fulfilled the primary outcome of sepsis (44 fulfilled the criteria for septic shock) and 22 (7%) died within 28 days. Table 1 details the baseline characteristics of the study cohort. Three quarters of the cases had infections of the respiratory tract, urinary tract and skin/soft-tissues, and 79 (24%) had positive blood cultures (Supplementary Table S1, Supplemental digital content 1, http://links.lww.com/EJEM/A192).
Two StO2 readings were recorded in 317 participants; five had only a T0 and one had only a T3 reading. The median time from ED presentation to T0 was 55 (interquartile range: 18–106) min and the median between T0 and T3 was 180 (174–190) min. At T0, the mean StO2 was 72 ± 10% in the sepsis group and 77 ± 9% in the nonsepsis group – mean difference 5% [95% confidence interval (CI): 2–7%; P < 0.0001]. At T3, the mean StO2 was 75 ± 9% in the sepsis group and 79 ± 8% in the nonsepsis group – mean difference 4% (95% CI: 2–6%; P = 0.0001). StO2 did not differ between the septic shock group and the nonshock sepsis group at either time point. There was an overall mean increase in StO2 between T0 and T3 (ΔStO2) of 2 ± 10%, but this did not differ between the nonsepsis, sepsis, septic shock or composite outcome (i.e. mortality/ICU admission >72 h) groups. Henceforth, the average of the two StO2 values for each participant (T0 and T3) will be reported. For the sepsis group, this was 74 ± 8% compared with 78 ± 7% in the nonsepsis group (P < 0.0001). For the composite outcome, the mean StO2 was 72 ± 9% compared with 77 ± 8% for those not fulfilling the outcome (P = 0.0001) (Fig. 2).
StO2 at T0 correlated with the peak SOFA score (Spearman’s ρ −0.27, P < 0.0001), (Supplementary Fig. S1, Supplemental digital content 1, http://links.lww.com/EJEM/A192), but there was no relationship with ΔStO2 (change from T0 to T3) (Supplementary Fig. S2, Supplemental digital content 1, http://links.lww.com/EJEM/A192). StO2 also correlated with the mortality in the ED sepsis score (−0.28, P < 0.0001). Despite differences in StO2 between the outcome groups, there was considerable overlap such that the ROC AUC for sepsis was 0.66 (95% CI: 0.60–0.72) and that for the composite outcome was 0.66 (0.58–0.75). Subgroup analyses on the basis of the source of infection (respiratory, urinary tract, nonurinary tract and those with positive blood cultures) did not yield markedly different results from the overall analysis (Supplementary Table S1, Supplemental digital content 1, http://links.lww.com/EJEM/A192).
Table 2 shows the univariate association of several predictor variables with outcome. In logistic regression analysis, StO2 less than 75% was associated with sepsis after adjustment for lactate, SBP, RR, GCS and Charlson Comorbidity Score (OR: 2.27; 95% CI: 1.23–4.17; P = 0.009). For composite mortality/ICU admission more than 72 h, StO2 less than 75% had OR 2.29 (1.07–4.88; P = 0.031), after adjusting for lactate, RR, SBP and GCS as covariates. Lactate was not measured in 44 cases, which were excluded from this analysis. Significant associations with outcomes were also observed with the variables SBP, GCS and RR, which comprise the ‘qSOFA’ severity prediction instrument developed as part of the sepsis 3 definition . We therefore carried out a post-hoc analysis to incorporate qSOFA into the model. The distribution of StO2 values by the qSOFA score is shown in Supplementary Figs S3 and S4 (Supplemental digital content 1, http://links.lww.com/EJEM/A192). The overall ROC AUC for the composite outcome for qSOFA was 0.69 (0.65–0.81) compared with that for StO2, 0.66 (0.58–0.75; P = 0.69). We also included lactate of at least 2 mmol/l as this was also found in sepsis-3 to confer mortality risk. The results show that StO2 less than 75% is associated with in-hospital mortality/ICU admission more than 72 h, independent of both qSOFA and lactate (Table 3). The ROC AUC for this model was 0.77 compared with 0.69 for qSOFA alone.
In a preplanned subgroup analysis, we assessed the utility for StO2 to predict subsequent deterioration in the SOFA score following admission. Of 173 cases with a SOFA score less than 1 on presentation, this increased by at least one point within the subsequent 72 h in 16 cases. ROC AUC for this outcome was 0.73 (0.66–0.79), with a sensitivity of 0.37 and a specificity of 0.76, respectively, at the cut-off point of StO2 less than 75%.
Finally, we also examined the utility of a StO2 more than 89% OR less than 75%. For sepsis, this had a ROC AUC of 0.63 (0.57–0.68) and for the composite outcome, it was 0.58 (0.52–0.63).
Sensitivity analyses were carried out first by excluding four cases in which the subsequent diagnosis was clearly found not to be because of infection (of whom one died). These cases were stroke, myocardial infarction with papillary muscle rupture, severe chronic anaemia and rhabdomyolysis secondary to drug overdose. Second, to assess for any confounding effect of delayed enrolment (for example the effect of early treatment on StO2 results), we also restricted the analysis to patients in whom T0 occurred within 180 min of arrival (n = 234). Neither of these sensitivity analyses yielded markedly different results.
Near-infrared spectroscopy-derived StO2 measured in the ED is associated with organ dysfunction and the diagnosis of sepsis, and with outcome in patients admitted to hospital with suspected infection. However, the predictive utility is insufficient for it to be used in isolation for this purpose. In logistic regression analysis, a low StO2 less than 75% is an independent risk factor for mortality and/or prolonged ICU admission. The finding that StO2 in conjunction with an elevated lactate and the qSOFA score predicts outcome is hypothesis generating, and requires independent prospective validation.
The strengths of the study include multicentre recruitment, pragmatic enrolment criteria, blinding of StO2 results and prespecified outcome measures corresponding to contemporary clinical definitions of sepsis. Limitations include restriction of recruitment to research personnel hours, resulting in convenience sampling, variation in the timing of enrolment in relation to ED presentation and enrolment of some participants in whom infection was subsequently excluded. However, sensitivity analysis indicated that these factors did not influence the overall results.
Previous investigators have examined the role of StO2 in the setting of sepsis in the ED. NIRS may be used to record StO2 as a single measure, as in our study, or the dynamic response to a vascular occlusion test . Shapiro et al.  performed a dynamic NIRS assessment among a cohort of patients with suspected sepsis. They found associations between low StO2 and organ failure/shock, with an impaired StO2 recovery slope after VOT being predictive of outcome. However, dynamic StO2 testing has practical limitations in the ED context, and is more suited to continuous monitoring in the theatre or the ICU setting. This has led to the development of small handheld devices designed to undertake measurement at a single time point, without a VOT. In a small, single-centre study, Vorwerk and Coates  found that persistent low StO2 ( < 75%) following resuscitation in septic shock was predictive of mortality. Leichtle et al.  found an association between low StO2 and the need for ICU admission among inpatients assessed for possible sepsis, but with limited predictive utility. Goerlich et al.  found that an abnormal StO2 measured at ED triage was predictive of sepsis diagnosis in combination with temperature, heart rate and RR among undifferentiated nontrauma presentations. In contrast, Goulet et al.  found no difference in StO2, measured at triage among patients presenting to ED, who had a final diagnosis of sepsis compared with those who did not. Finally, in a study carried out in an acute cancer centre, Bazerbashi et al.  found that an abnormally low StO2 was predictive of the need for ICU admission.
It is important to recognize the patient at risk of developing organ failure among those presenting with features of infection. Alterations of microcirculation are an important component of the pathogenesis of sepsis. However, the assessment of microcirculatory dysfunction in the clinical setting is difficult. NIRS offers a potential solution in being a bedside test that is noninvasive and easily repeatable. Both abnormally low StO2 ( < 75%) and a high level ( > 89%) have been suggested to be associated with sepsis . Our results are broadly consistent with previous works, which support an association between abnormal StO2 and organ failure/outcome in sepsis. However, the ability of StO2 to differentiate between clinical outcome groups of interest is limited, and NIRS cannot be recommended in isolation for this purpose on the basis of our findings.
Our study focused only on single StO2 measurements without dynamic testing . We also cannot make inferences on the utility of other perfusion or microcirculatory assessment tools . We focused on StO2 for patient classification and outcome prediction; however, the potential for StO2 as a resuscitation target in sepsis remains a possible future avenue for research. However, a small pilot study in ICU patients found no difference in mortality with a resuscitation target of StO2 of at least 80% . Finally, although StO2 has limited clinical utility as a stand-alone test, its incorporation into a risk score along with other commonly assessed predictor variables could be considered in a future study.
NIRS-derived StO2 is associated with sepsis diagnosis, organ failure and outcome among patients admitted to hospital from the ED with infection. However, it has limited predictive utility to differentiate between clinical outcome groups of interest, and cannot be recommended for this purpose.
The authors acknowledge the substantial contribution of the following individuals to this work: Simon Brown, Niall Henry, Ellen Macdonald, Sophie Damianopoulos, Jonathon Burcham and the research nurses at the three participating hospitals.
The ARISTOS study was investigator-initiated and funded by the Emergency Medicine Foundation, The Prince Charles Hospital Foundation and by the participating institutions. Dr Ho is funded by the Raine Medical Research Foundation and WA Health through the Raine Clinical Research fellowship.
Conflicts of interest
There are no conflicts of interest.
1. Singer M, Deutschman CS, Warren Seymour C, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis
and septic shock
-3). JAMA 2016; 315:801–810.
2. Rhodes A, Evans LE, Alhazzani W, Levy MM, Antonelli M, Ferrer R, et al. Surviving Sepsis
Campaign: International guidelines for management of sepsis
and septic shock
: 2016 Crit Care Med 2017; 45:1–67.
3. Hampton DA, Schreiber MA. Near infrared spectroscopy
: clinical and research uses. Transfusion 2013; 53Suppl 1525–585.
4. Neto AS, Pereira VGM, Manetta JA, Esposito DC, Schultz MJ. Association between static and dynamic thenar near-infrared spectroscopy
and mortality in patients with sepsis
: as systematic review and meta-analysis. J Trauma Acute Care Surg 2014; 76:226–233.
5. Macdonald SPJ, Brown SGA. Near-infrared spectroscopy
in the assessment of suspected sepsis
in the emergency department. Emerg Med J 2015; 32:404–408.
6. Crawford J, Otero R, Rivers EP, Goldsmith D. Near-infrared spectroscopy
as a potential surrogate for mixed venous oxygen saturation for evaluation of patients with hemodynamic derangements. Crit Care 2008; 12Suppl 2P69.
7. Nardi O, Gonzalez H, Fayssoil A, Annane D. Masseter muscle oxygen saturation is associated with central venous oxygen saturation in patients with severe sepsis
. J Clin Monit Comput 2010; 24:289–293.
8. Mesquida J, Masip J, Gill G, Artigas A, Baigorri F. Thenar oxygen saturation measured by near infrared spectroscopy
as a noninvasive predictor of low central venous oxygen saturation in septic
patients. Intensive Care Med 2009; 35:1106–1109.
9. Soga T, Sakatani K, Yagi T, Kawamorita T, Yoshino A. The relationship between hyperlactatemia and microcirculation in the thenar eminence as measured using near-infrared spectroscopy
in patients with sepsis
. Emerg Med J 2014; 31:654–658.
10. Goerlich CE, Wade CE, McCarthy JJ, Holcomb JB, Moore LJ. Validation of sepsis
screening tool using StO2 in emergency department patients. J Surg Res 2014; 190:270–275.
11. Shapiro NI, Arnold R, Sherwin R, O’Connor J, Najarro G, Singh S, et al. The association of near-infrared spectroscopy
-derived tissue oxygen measurements with sepsis
syndromes, organ dysfunction and mortality in emergency department patients with sepsis
. Crit Care 2011; 15:R223.
12. Goulet H, Andre S, Der Sahakian G, Freund Y, Khelifi G, Claessens YE, et al. Accuracy of oxygen tissue saturation values in assessing severity in patients with sepsis
admitted to emergency departments. Eur J Emerg Med 2014; 21:266–271.
13. Vorwerk C, Coates TJ. The prognostic value of tissue oxygen saturation in emergency department patients with severe sepsis
and septic shock
. Emerg Med J 2012; 29:699–703.
14. Leichte SW, Kaoutzanis C, Brandt MM, Welch KB, Purtill MA. Tissue oxygen saturation for the risk stratification of septic
patients. J Crit Care 2013; 28:1111–1111–e5e1–1111–e5.
15. Bazerbashi H, Merriman KW, Toale KM, Chaftari P, Carreras MTC, et al. Low tissue oxygen saturation at emergency center triage is predictive of intensive care unit admission. J Crit Care 2014; 29:775–779.
16. Levy MM, Fink MP, Marshall JC, Angus D, Cook D, Cohen J, et al. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis
Definitions Conference. Intensive Care Med 2013; 29:530–538.
17. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method for classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40:373–383.
18. Hasanin A, Mukhtar A, Nassar H. Perfusion indices revisited. J Int Care 2017; 5:24.
19. Nardi O, Polito A, Aboab J, Gwenhael C, Virgine M, Clair B, et al. StO2 guided resuscitation in subjects with severe sepsis
or septic shock
: a pilot trial. J Clin Monit Comput 2013; 27:215–221.
Keywords:Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.
emergency service; hospital; near-infrared; sepsis; septic; shock; spectroscopy