The Third International Consensus defines sepsis as a ‘life-threatening organ dysfunction caused by a dysregulated host response to infection’ (Singer et al, 2016). Sepsis is a leading cause of death in the UK with 245 000 cases reported annually with a mortality rate of 20.3% (close to 50 000 deaths); globally, there are 49 million cases with 11 million deaths per year (UK Sepsis Trust, 2022).
Sepsis causes a multitude of pathological cascades, which lead to end-organ damage and homeostatic imbalances (Singer et al, 2016). These cascades lead to increased lactate production.
During the progression of sepsis, cardiovascular dysfunction because of septic shock begins to lead to systemic hypoperfusion and respiratory dysfunction (Gyawali et al, 2019). Tissue permeability caused by the cardiovascular dysfunction around the lungs leads to pulmonary oedema causing a ventilation–perfusion mismatch, resulting in hypoxia (Gotts and Matthay, 2016).
The combination of respiratory and cardiovascular impairment lowers the blood pressure and causes hypoxia, leading to tissue hypoperfusion and organ dysfunction (Gyawali et al, 2019).
As a result of the tissue hypoperfusion, cells are starved of oxygen and become hypoxic, so are forced to respire using anaerobic metabolism. The by-product of this metabolism is lactate (Seymour et al, 2016). During global hypoxia, a large amount of lactate is produced, which creates a metabolic acidosis known as lactic acidosis (Singer et al, 2016).
Studies into reducing mortality have shown that rapid diagnosis and rapid treatment of sepsis can significantly reduce mortality (Seymour et al, 2017). Several screening tools such as the National Early Warning Score (NEWS) 2 (Royal College of Physicians, 2017) and the quick sequential organ failure assessment (qSOFA) score (Seymour et al, 2016) have been adopted in both prehospital and hospital settings and have led to effective recognition of sepsis (Goulden et al, 2018).
In the emergency setting, capnography is used to measure expired carbon dioxide (CO2). CO2 is quantified in two ways which are often combined.
Capnography is an umbrella term. First, it covers capnometry, which measures peak expiratory CO2 and is measured as end-tidal CO2 (EtCO2) and recorded in either mmHg or kPa. Second, it includes waveform capnography, which displays a waveform showing expired CO2 over each breath (Valente, 2010).
EtCO2 can be affected by several pathologies and provides an insight into a patient's metabolic, ventilation and cardiovascular status as all three affect the release of CO2 (Valente, 2010).
The pathology of most interest in sepsis is metabolic change. EtCO2 indicates the body's metabolic state because it is correlated to bicarbonate (HCO3) and arterial partial pressure of CO2 (PaCO2) (Pishbin et al, 2015). In metabolic acidosis, increases in respiratory rate and tidal volume (Flenady et al, 2017) cause increased expiration of CO2, leading to decreased CO2 pressure and reduced EtCO2 readings (Wiryana et al, 2017). A link to sepsis forms here as an increase in lactate levels can lead to metabolic acidosis (Kamel et al, 2020) and therefore EtCO2 could be useful in the identification of sepsis.
Methods
A systematic search was conducted to find strong primary quantitative research to determine to what extent low EtCO2 is a predictor of sepsis and if this has diagnostic value.
To achieve this, two databases were searched between 15 February and 3 March 2021. Both CINAHL Plus and Medline were accessed through EBSCOhost. CINAHL Plus was chosen for its large literature pool as it covers over 4600 journals relating to allied health and nursing (Bhat and Takoor, 2014), whereas Medline was chosen for its wide literature base and relevance, and because it is a useful and widely accessible database for healthcare (Dunn et al, 2017).
The search was conducted on EBSCOhost, a database search platform that allows MEDLINE and CINAHL to be searched simultaneously and automatically omits duplicates in the two databases. EBSCOhost allows for the use of an advanced search and supports the use of boolean phrase searching and truncation (Aliyu, 2017). These terms allowed the search to include related words such as EtCO2 or end-tidal carbon dioxide or capnography to obtain all relevant literature. Broad keywords included EtCO2, end-tidal carbon dioxide, capnography, sepsis, septic shock, identification, screening, detection, prediction and diagnosis.
The results of the literature search were screened first by title then abstract and, finally, the full text was evaluated against Critical Appraisal Skills Programme (2018) checklists.
Articles included had to be primary research available in full text, published after 2006 and use EtCO2 for the identification of sepsis; they may originate from any country and include any patient age groups. Animal studies, using EtCO2 as a treatment guide or those where sepsis was not mentioned in the patient group, were excluded. This screening process is documented in the PRISMA diagram (Figure 1) and allowed the literature to be filtered to find the most appropriate studies to address the review question.
Results
The database search identified 44 papers (Figure 1). Three duplicates had to be manually removed and 41 papers were screened against the inclusion criteria with 20 excluded at the title and 10 at the abstract stages. Eleven papers were full text, which were reviewed and critiqued, and four were excluded based on relevance and level of rigour. Seven papers were included in the final review (Table 1).
| Author, year, title | Methods | Results | Conclusion |
|---|---|---|---|
|
McGillicuddy et al (2009)
|
A pilot prospective observational cohort study in an emergency department in Israel recruiting 97 patients aged ≥18 years with a fever. Recording their EtCO2, SOFA scores and lactate to assess for the relationship and predictive value of EtCO2 related to SOFA scores and lactate level | EtCO2 of <35 has a sensitivity of 0.73 (95% CI (0.56–0.85)) and specificity 0.50 (0.38–0.62) in predicting SOFA scores. EtCO2 <35 had a sensitivity of 0.60 (0.22–0.88) and specificity of 0.42 (0.32–0.52) in predicting lactate (4 mmol/l) | There was a statistically significant correlation between EtCO2 and SOFA score and serum lactate levels; however, operating characteristics of EtCO2 in predicting lactic acidosis or end-organ dysfunction in this patient population might not be reliable enough for clinical decision-making |
|
Hunter et al (2013)
|
A prospective observational study at Orlando Regional Medical Center ED in the United States recruiting patients aged ≥18 years with suspected infections and meeting sepsis criteria. 201 patients were included and data on lactate, EtCO2 and outcomes were recorded, and the association between EtCO2 and in-hospital mortality and the correlation between EtCO2 and serum lactate calculated | There was a significant inverse relationship between EtCO2 and lactate in all categories, with correlation coefficients of −0.421 (P<0.001) in the sepsis group, −0.597 (P<0.001) in the severe sepsis group and −0.482 (P=0.011) in septic shock. The areas under the curve for EtCO2 and mortality were 0.60 (95% CI (0.37–0.83)) for sepsis, 0.67 (95% CI (0.46–0.88)) for severe sepsis and 0.78 (95% CI (0.59-0.96)) for septic shock | There was a significant association between EtCO2 concentration and in-hospital mortality in emergency department patients with suspected sepsis. Also a significant inverse relationship between lactate and EtCO2 |
|
Hunter et al (2016)
|
A prospective cohort study using emergency medical services in Florida, Orange County, in the United States (US), recruiting patients aged ≥18 years who met the county's sepsis protocol. 298 patients were included and prehospital observations including EtCO2 were recorded and compared with in-hospital diagnosis and blood test results | There was a significant correlation between EtCO2 and HCO3 levels, with a correlation coefficient of 0.415 (P<0.001). EtCO2 had a higher power to predict sepsis and mortality than the other variables; the area under the ROC curve predicting sepsis was 0.99 for EtCO2 (95% CI (0.99–1.00); P<0.001). A sepsis protocol incorporating ≥2 SIRS criteria and EtCO2 ≤25 mmHg in adults with suspected infection predicted sepsis (69% sensitivity, 67% specificity), severe sepsis (90% sensitivity, 58% specificity), and mortality (76% sensitivity, 46% specificity) | Among all collected prehospital vital signs, low EtCO2 concentration had the best predictive value for sepsis |
|
Taghizadieh et al (2016)
|
A descriptive-analytic study was conducted in an emergency department in Iran. Recruited 262 patients (80 sepsis patients) of all ages presenting with an ABG confirmed metabolic acidosis. The study explored the link between HCO3 and EtCO2 | A significant direct linear relationship between the EtCO2 and the HCO3 level of arterial blood in patients with sepsis (P<0.001 and r=0.431) | Results show that performing capnography is recommended for the detection of metabolic acidosis in awake patients suspected of acid-base disorders. According to the results of this study, capnography can be used for the primary diagnosis of metabolic acidosis in spontaneously breathing patients |
|
Hunter et al (2018)
|
A retrospective cohort study involving EMS in Florida (US) that recruited 287 patients aged ≥18 years who met the county's sepsis protocol. The primary outcome was the relationship between EtCO2 and qSOFA scores and in-hospital mortality. The secondary outcome was a diagnosis of sepsis or severe sepsis upon hospital admission. | Sensitivity and specificity for EtCO2 as a mortality predictor was higher, 80% (95% CI (59–92)) versus 68% (95% CI (46–84)), and 42% (95% CI (36–48)) versus 40% (95% CI (34–46)), respectively, than qSOFA score. The ROC curve predicting sepsis was 0.66 for EtCO2 (95% CI (0.59–0.72); p<0.001) and 0.61 for qSOFA (95% CI (0.54-0.68); p=0.002) | By comparison of ROC curves, EtCO2 had a higher discriminatory power to predict mortality, sepsis and severe sepsis than qSOFA. Both non-invasive measures were easily obtainable by prehospital personnel, with EtCO2 performing slightly better as an outcome predictor |
|
Weiss et al (2019)
|
A retrospective cohort study was conducted in an Albuquerque (US) EMS department. The study recruited 351 patients of all ages and compared EtCO2 to lactate, mortality, admission to hospital, and sepsis diagnosis | EtCO2 (<25) predicted a positive lactate (>4.0 mmol/l) with 76% accuracy, 63% specificity and 80% sensitivity (OR=26). 27% of patients with a positive EtCO2 and 29% of patients with negative EtCO2 had a positive primary outcome of a source of infection found | EtCO2 was effective at predicting first hospital lactate levels but did not reliably predict the primary outcome of infection or admission to hospital |
|
Cully et al (2020)
|
A prospective study at a level 1 trauma tertiary care children's hospital that recruited 69 patients aged ≥30 days to ≤21 years who met the criteria for sepsis, severe sepsis or septic shock. EtCO2 and lactate were measured as part of the sepsis bundle, and the association between them recorded and compared with the diagnosis of sepsis | There was a significant inverse relationship between EtCO2 and lactate with a correlation coefficient of −0.34 (P=0.005). There was a significant difference between EtCO2 (P=0.004) and lactate (P<0.001) in suspected sepsis, severe sepsis and septic shock. An EtCO2 cut-off point of 30 mmHg correlated with a sensitivity of 93% and specificity of 32% for predicting lactate | A significant inverse relationship was found between EtCO2 and lactate in children presenting to a paediatric emergency department with suspected sepsis. A lower EtCO2 was predictive of severe disease |
ED: emergency department; EMS: emergency medical services; EtCO2: end-tidal carbon dioxide; HCO3: bicarbonate; qSOFA: quick sequential organ failure assessment; ROC: receiver operating characteristic; SOFA: sequential organ failure assessment
Discussion
As it is a life-threatening condition causing major organ dysfunction to develop (McGillicuddy et al, 2009), sepsis requires rapid treatment to reduce mortality. Many large-scale studies have shown that effective rapid treatment of sepsis can reduce mortality by up to 16% in adults (Hunter et al, 2013) and by up to five times in children (Cully et al, 2020).
The main issue facing clinicians in the management of sepsis is its recognition. The Third International Consensus Definitions for Sepsis document has highlighted the use of the qSOFA and SOFA scores as sepsis identification tools (Singer et al, 2016) and measurement of serum lactate is an accepted way of assessing sepsis and its severity (Hunter et al, 2016).
EtCO2 measurement has recently gained traction for its use in identifying other metabolic-based conditions such as diabetic ketoacidosis (Taghizadieh et al, 2016). EtCO2 monitoring is deemed a non-invasive measurement that can be easily added to traditional vital sign assessments (Weiss et al, 2019). As it can provide real-time analysis of a patient's perfusion, ventilatory and basal metabolic status (Hunter et al, 2013), it may be useful in the diagnosis of sepsis.
This review identified three main themes that are explored below: EtCO2 and lactate; EtCO2 in comparison and in conjunction with sepsis screening tools; and EtCO2 as a predictor of mortality.
EtCO2 and lactate
Through an examination of the relationship between lactate in sepsis and EtCO2, the association between EtCO2 and sepsis can be explored.
Lactate is considered a key indicator of sepsis and is linked to the severity of the condition (Hunter et al, 2016).
To quantify sepsis severity, it can be split into the outdated categories of sepsis, severe sepsis and septic shock as outlined at the 2001 International Sepsis Definitions Conference (Levy et al, 2003). Hunter et al (2016) found that mean lactate levels in sepsis are 1.9 mmol/l (1.9–2.1), severe sepsis 5.4 mmol/l (4.8–6.2) and septic shock lactate 3.2 mmol/l (2.8–3.5). This is supported by Cully et al (2020), who explored this link in paediatric patients and found mean lactate in sepsis to be 1.9±1.1 mmol, 2.8±1.3 mmol for severe sepsis and 3.4±1.8 mmol for septic shock.
As a link between lactate and sepsis has been established, the link between lactate and reduced EtCO2 can be explored. There is already a rational physiological link (Cully et al, 2020), and evidence appears to support this.
Hunter et al (2018) compared patients' lactate levels to their EtCO2 and found a mean EtCO2 of 28 mmHg (27–29 mmHg) in sepsis, 19 mmHg (18–22 mmHg) in severe sepsis, and 25 mmHg (16–24 mmHg) in septic shock, indicating EtCO2 was reduced in patients with a high lactate level and a possible inverse correlation. Hunter et al (2013) explored this correlation and found correlation coefficients in sepsis, severe sepsis and septic shock were −0.421 (P=0.001), −0.597 (P=0.001) and −0.482 (P=0.011) respectively, indicating a moderate but significant inverse correlation.
To examine if EtCO2 can be accurately employed as a predictor of lactate levels, its sensitivity and specificity need to be assessed.
Cully et al (2020) and Weiss et al (2019) state sensitivity as 97% with a cut-off of <30 mmHg and 80% with a cut-off of <25 mmHg respectively demonstrating a strong ability to identify high lactate. They also stated specificity was 31% and 63% respectively, which suggests that a lower EtCO2 cut-off may provide a higher specificity.
Therefore, there could be some clinical application of EtCO2 as a tool for identifying high lactate and therefore sepsis in the prehospital setting or where measuring arterial blood gas (ABG) may take time; periods of up to 172 minutes after arrival in the emergency department (ED) have been noted (Hunter et al, 2013).
As EtCO2 is quick to apply and can be included in standard observations (Weiss et al, 2019), it may prove to be a beneficial tool in reducing identification times and therefore time to treatment.
A further metabolic link between EtCO2 and sepsis can be made by looking at HCO3. Taghizadieh et al (2016) found a moderate but significant relationship of r=0.431 for EtCO2 predicting HCO3 in patients with sepsis and concluded that EtCO2 could be used to make a primary diagnosis of sepsis while awaiting ABG results.
EtCO2 in comparison and conjunction with sepsis screening tools
To evaluate the effectiveness of EtCO2 at predicting sepsis, it would be appropriate to compare it to validated sepsis screening tools—the qSOFA and the systemic inflammatory response syndrome (SIRS) criteria.
qSOFA was introduced in the Third International Consensus Definitions for Sepsis (Singer et al, 2016), and uses three measures to predict sepsis and sepsis mortality. Scores of ≥2 indicate sepsis and an increased risk of mortality.
In one study, a qSOFA of ≥2 had an area under the receiver operating characteristic (ROC) curve of 0.61 for predicting sepsis and 0.69 for severe sepsis, and a sensitivity and specificity of 68% and 40% respectively in predicting mortality (Hunter et al, 2018). In comparison, in the same study, an EtCO2 of <25 mmHg had an area under the ROC curve of 0.66 for predicting sepsis and 0.78 for severe sepsis, and a sensitivity and specificity of 80% and 42% respectively in predicting mortality (Hunter et al, 2018).
This indicates that an EtCO2 of <25 mmHg is a more effective predictor of both sepsis and severe sepsis and mortality than a score of ≥2 or more on the qSOFA criteria.
This study also investigated combining qSOFA and EtCO2. This combined approach found an area under the ROC curve of 0.68 for predicting sepsis and 0.81 for severe sepsis (Hunter et al, 2018), which suggests the combined approach has a higher predictive capability than each of the assessments individually.
Systemic inflammatory response syndrome is a syndrome caused by severe stress on the body such as infection or trauma. In the 2001 International Sepsis Definitions Conference (Levy et al, 2003), SIRS was considered a key component of sepsis, and the SIRS criteria were considered key in the diagnosis of sepsis. Recent changes in the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis–3) (Singer et al, 2016), mean that SIRS is no longer recognised as a syndrome involved in sepsis; however, the definition still recognises the diagnostic capability of the SIRS criteria for infection.
A study by Weiss et al (2019) explored the ability of the SIRS criteria (≥2) compared to EtCO2 (≤25 mmHg) at predicting infection and/or death. They found that SIRS criteria provided a statistically significant (P<0.05) OR of 3.5 in identifying a source of infection and a significant OR of 3.9 in predicting hospital admission, whereas EtCO2 did not provide a statistically significant OR (0.9) in identifying infection or a significant OR (0.8) in predicting hospital admission.
SIRS, therefore, is a more effective tool in predicting the likelihood of infection and hospital admission than EtCO2. However, Hunter et al (2016) found that a combined screening tool using the SIRS criteria (≥2) and EtCO2 (≤25 mmHg) had a 69% sensitivity and 67% specificity in predicting sepsis, and a 90% sensitivity and 58% specificity in predicting severe sepsis. This indicates that EtCO2, when combined with SIRS, may have a role to play in sepsis identification.
EtCO2 as a predictor of mortality
A good indicator of how useful a diagnostic tool could be is its ability to predict mortality. Predicting outcomes can help clinicians identify high-risk patients and reduce their risk of dying (Hunter et al, 2013).
Lactate is an accepted predictor of mortality in sepsis and, with the link between EtCO2 and lactate established, EtCO2 has the potential to be an effective predictor of mortality.
Hunter et al (2013) explored the correlation between lactate and mortality and EtCO2 and mortality using ROC curves; the area under the curve (AUC) of 0.75 for lactate and mortality supports the suggestion that lactate is an effective tool for predicting mortality. A similar AUC of 0.73 for EtCO2 and mortality indicates a similar ability of EtCO2 to predict mortality. The researchers do not mention a particular cut-off for predicting mortality but instead, state that a higher lactate or a lower EtCO2 is predictive of mortality.
Further research by Hunter et al (2018) quantified a cut-off to predict mortality and found a cut-off of ≤25 mmHg predicts mortality with a sensitivity of 80% and specificity of 42%.
The use of EtCO2 to predict mortality is also supported by Weiss et al (2019), who found EtCO2 alone to have a significant positive predictive value for mortality with a statistically significant OR of 2.8. All studies found in this review support the use of EtCO2 as a predictor of mortality and agree on its benefits of identifying higher-risk patients to focus treatment.
Conclusion
The use of EtCO2 is becoming widespread in healthcare and the research presented in this review largely supports using EtCO2 as an effective tool at diagnosing sepsis. EtCO2 has a strong physiological link with sepsis; this review also found strong evidence supporting that EtCO2 is inversely correlated with lactate levels, which have a recognised diagnostic value for sepsis. It can therefore be concluded that EtCO2 is an effective diagnostic tool for sepsis, although larger trials are needed.
Furthermore, in a comparison with recognised sepsis screening tools, EtCO2 has proved to be a more effective indicator of sepsis than the recognised and recommended qSOFA score and, when used in conjunction with qSOFA, creates more effective diagnostic criteria. Similarly, when EtCO2 is compared with SIRS, a combined approach provides a higher sensitivity and specificity than SIRS alone.
The review determined that EtCO2 is also an indicator of disease severity with a strong ability to predict mortality, suggesting that EtCO2 may play a role in early sepsis triage and clinical decision-making for the most acutely unwell patients.
The evidence presented by this review unanimously supports that EtCO2 is an effective tool in diagnosing sepsis, is comparable or exceeds the diagnostic capabilities of validated sepsis diagnosis tools and is easy to apply and monitor.
The research proposes several cut-offs as to when EtCO2 is diagnostic for sepsis. The most common cut-off with the most supportive evidence is ≤25 mmHg (3.3 kPa), so it can be suggested that an EtCO2 of ≤25 mmHg in patients with a suspected infection is diagnostic of sepsis. This can therefore be used to increase the speed and accuracy of sepsis diagnosis thus potentially reducing sepsis mortality.
Recommendations
There are three recommendations for further research and one for current practice, focusing on future impact to practice in the UK.
The research in this review is primarily US based and focuses on US ED and prehospital systems, which differ from those in the UK in terms of the way they are structured, their scope of practice and training. Therefore, it is recommended that further large-scale research be conducted into the use of EtCO2 in the diagnosis of sepsis by UK emergency departments and prehospital providers. The UK Sepsis Trust has its own set of validated sepsis screening tools for ED and prehospital use (Nutbeam and Daniels, 2021).
It is further recommended that a prospective cohort trial be conducted to investigate and compare the effectiveness of the UK sepsis screening tools and EtCO2 in the diagnosis of sepsis and to evaluate the benefits of combining them.
The research mentioned in this study focuses on initial EtCO2. The studies also mention a potential use of EtCO2 in the ongoing management of sepsis and using EtCO2 to monitor the effectiveness of treatment. Therefore, this review suggests that research be carried out into using EtCO2 in the ongoing care of patients to guide treatment and ensure positive outcomes.
For current practice, although research largely supports the use of EtCO2 in diagnosing sepsis, larger-scale studies in the UK need to be completed before it can be recommended as a screening tool in practice.