Background

Medical attention to out-of-hospital cardiac arrest (OHCA) entails a variety of special challenges, besides the inherent ones when coping with this type of pathology, in any context. One of them would be the lack of exactness when estimating the time of the cardiac arrest (CA), which may be of utter relevance. No-flow time is probably one of the key factors in the neurological outcome [1,2,3]. Some clinical signs have been considered to estimate the cardiac arrest (CA) onset time but they are not reliable enough to influence on the decision to initiate, continue or stop the cardiopulmonary resuscitation (CPR).

Along with the cardiac arrest onset and first monitored rhythm, it seems that the metabolic status during CPR would be another important factor with a relevant impact on the probability of survival and on the neurological outcome [4].

Multiple responses have been sought in this regard, analysing factors that could be related to the neurological outcome and, therefore, that could lead to determine whether to maintain or to stop all resuscitation efforts. In most cases, such studies analyse the patient’s status after the recovery of spontaneous circulation (ROSC), either on arrival at the emergency department or at the critical-care unit in the hospital. Thus, the interaction between the forementioned neurological outcome and certain aspects of the CA has been evaluated, [5,6,7,8,9] such as the first monitored rhythm, the presence of first responders or the duration of resuscitation efforts, in some cases using scales and other predictive tools [10, 11].

Several blood metabolite concentrations were evaluated while looking for possible predictive tools and obtaining positive results. Recently, an important contribution was made in this search for predictors to associate blood metabolites and the neurological outcome [12]. Amongst them, the relationship of variables, such as PaCO2 or lactate, and the chance of survival with a good neurological outcome are remarkable. Several studies have demonstrated the statistical association between hypocapnia [13, 14] and neurological damage. Likewise, high levels of lactate [15, 16] at 6 and 12 h after resuscitation can be significant predictors of poor neurological outcome, but it does not seem to operate in earlier stages [17]. The relationship with pH and other blood-gas values has also been assessed but almost always after resuscitation [18,19,20]. Measurements after the transference to the hospital can be accurate predictors but are not suitable for the emergency medical service (EMS) team and are influenced by advanced CPR.

In Spain and some other countries where doctors and nurses are also working in mobile advanced life support (ALS) units, the common course of action is to do CPR on the field and transfer the patient to the hospital if ROSC is achieved or declare him/her dead at the scene in the opposite case. Only in very specific situations, such as patients who recover and lose spontaneous circulation several times, or if any other mechanical alternative is considered feasible, may the patient be transferred with ongoing CPR.

This study aims at analysing the metabolic situation at the start of advanced CPR at the scene and its relation with the neurological condition after 1 month. Our group has wide experience with the use of point-of-care analysis on the field. Since there is often very limited data regarding a patient’s previous medical history and time of arrest, on-scene lab studies (at the beginning of CPR) could be useful for decision-making and provide some kind of metabolic watch of the patient and therefore become a tool to determine the likelihood of successful resuscitation [21, 22].

We aim to study the association between the analytical variables at the start of advanced CPR and two main outcomes: ROSC at arrival to the hospital and the neurological condition of the patients 30 days after the event, assessed by the Cerebral Performance Categories (CPC) scale, considering intact neurological recovery grades I and II.

Methods

Study design and setting

This was a prospective observational cohort study of all non-traumatic OHCA in patients older than 17 years assisted between 2012 and 2017 by the EMS SAMUR-PC. EMS participants collected the data at the scene following the Utstein style. Epidemiological variables were recorded (age, sex, rhythm of cardiac arrest onset, witnessed cardiac arrest, previous manoeuvres by first responders/bystanders).

As stated in the procedures of this EMS, [23] in every attended cardiac arrest, a venous blood sample was taken upon initially obtaining venous access (in the first 90 s) and analysed on the scene through the point-of-care EPOC® (Epocal Inc., Ottawa, Canada) device. This venous access was almost always achieved in the arms. If it was not possible, intraosseous access was achieved. These initial blood test results (pH, pCO2, HCO3−, base excess [BE], Na+, K+, Ca2+ and lactate), prior to the administration of any drug, were analysed as independent variables in this study.

Cases in which CPR was not initiated, the patient was under aged or blood analysis could not be performed were excluded.

SAMUR-PC is an EMS whose scope of action is developed in all public settings of the city of Madrid (Spain), with a population of 3.2 million inhabitants and 140,000 dispatches per year. A physician and a nurse are on board all mobile ALS units. In every suspected very severe patient, a mobile ALS unit is dispatched, as well as the chief doctor and the head nurse.

Our service is enrolled in an international study on ECPR. The four cases included in that study have not been considered for the present one. When ROSC is not achieved after efforts according to recommendations—and the patient would otherwise be declared dead—he/she could be transferred to the hospital in asystole if meeting the inclusion criteria for uncontrolled non-heart beating donation. In the present study, these patients are included in the no-ROSC group.

Study endpoints

The dependent variables were ROSC upon arrival at the hospital and intact neurological survival (CPC I-II) at 30 days. CPC scale was performed by in-person interview by the responsible physician of the patient at the hospital.

Statistical analysis

The description of quantitative variables was performed with central tendency and dispersion indices based on rankings (mean ± standard deviation). The description of categorical variables was performed with absolute and relative frequencies in percentages.

Proportions were compared with Pearson’s chi-squared tests to analyse the relationships between categorical variables. Quantitative variables were analysed with the Student’s t test.

In the first instance, a univariate analysis was carried out using simple binary logistic regression between each independent variable and the dependent variables.

To assess the relationship of the independent variables associated with the dependent variables, we used multivariate binary logistic regression models, one for each of the variables (pH, pCO2, HCO3−, BE, Na+, K+, Ca2+ and lactate). Each model was evaluated separately for each independent variable of interest and was adjusted with the following covariables, selected based on their relationship with the dependent variables in the scientific literature [3]: age, sex, first monitored rhythm, witnessed CA, previous manoeuvres and the interactions of these covariates with each independent variable of interest. Estimative, multivariate, binary, logistic regressions were performed, selecting the final model by using the backward procedure with the criterion of statistical significance to remove covariates. The likelihood ratio test was used to analyse the overall statistical significance of the models. The Wald test was used for the individual statistical significance of the predictors. The magnitude of the effect of each independent variable was expressed with the odds ratio (OR) and its 95% confidence interval (CI). In continuous variables, it was analysed using intervals of increment or decrement and OR was expressed per unit of analyte. Internal validation of the final multivariate models was carried out by dividing the sample by time criteria, using the most recent 80% of the sample to estimate the models and the oldest 20% of the sample to validate the models. Values of p < 0.05 were considered significant. The statistical treatment was performed with the SPSS, version 18 (SPSS Inc., Chicago, IL, USA) statistical package.

Results

A total of 1678 OHCAs were consecutively attended by the EMS. One hundred and twenty-six cases (7.5%) were lost due to not being able to obtain the blood analytical data due to an ambient temperature that made the device unusable. Thus, the study was performed on 1552 records (Fig. 1). Only 3% of them occurred at home/residence (the rest of them were in the street or public buildings, including workplace, gyms and educational institutions) and 91% were witnessed. The mean response time (from incoming call to arrival of the first vehicle) was 8 min and 14 s.

Fig. 1
figure 1

Study scenario. Central green boxes summarise data from all the records included. CPC I-II percentages are out of all patients, not out of patients who recovered spontaneous circulation

Full size image

Epidemiological characteristics of the population studied are shown in Table 1. In 50.2% of cases, first responders/bystanders performed CPR prior to EMS arrival. ROSC was achieved in 907 patients (58.44%) patients; in the subgroup of shockable rhythms, ROSC was achieved in 70.4% of cases. Table 2 shows the association of ROSC and good neurological recovery with other epidemiological variables. In 24.68% of 1552 patients, there was no neurological damage or only minimal sequelae after 1 month (CPC I-II); in the subgroup of shockable rhythms, this percentage was 43.2%.

Table 1 Epidemiological values of the study population (n = 1552)
Full size table
Table 2 Association of ROSC (n = 907) and good neurological recovery CPC I-II (n = 383) with other epidemiological variables
Full size table

In the univariate analysis, a significant relationship was found between non-recovery of spontaneous circulation (no-ROSC) and lower pH and higher levels of pCO2, potassium and calcium (Table 3). Likewise, there was a significant relationship between poor neurological recovery (no-CPC I-II) and a decrease in pH, HCO3− and BE and an increase in pCO2 and potassium (Table 3).

Table 3 Comparison of analytical values according to ROSC and neurological recovery CPC I-II (univariate analysis)
Full size table

A significant relationship between non-recovery of spontaneous circulation (no-ROSC) and low pH and high pCO2 and potassium levels was found in the multivariate analysis using intervals of increment or decrement. The internal validation of the models was positive for pH and pCO2 and negative for potassium (Table 4). In the same way, the multivariate analysis showed a significant relationship between neurological non-recovery (no-CPC I-II) and low pH, HCO3− and BE and high pCO2 and potassium levels. The internal validation of the models was positive for pH, pCO2, BE and potassium and negative for HCO3− (Table 4).

Table 4 Analysis with multivariate binary logistic regression for ROSC and good neurological recovery (CPC I-II)
Full size table

We continuously assessed the relationship between the values of the different analytes and the outcome variables (Fig. 2), after noticing a proportional relationship between the levels of each predictor and those variables. This relationship was very close when referring to the values of pH and pCO2, both for ROSC and for good neurological recovery.

Fig. 2
figure 2

a Relationship between the percentages of recovery of spontaneous circulation (ROSC) and 30-day survival with good neurological outcome (CPC I-II) and the values of pH and pCO2 expressed in ranges evaluated in the multivariate binary logistic regression. b Relationship between the percentages of survival with good neurological outcome (CPC I-II) and the values of base excess (BE) and K+ expressed in ranges evaluated in the multivariate binary logistic regression. Venous blood-gas variables, including alterations in blood potassium, are associated with neurological outcomes. Low pH, a raised pCO2 and a high base deficit, as well as either very low or high blood concentration of potassium, were associated with worse outcome

Full size image

Discussion

The importance of these analyses lies in the fact that the study was carried out in a setting in which all patients were attended by emergency teams with physician, nurse and several emergency technicians (paramedics), usually helped by an additional chief physician and/or nurse, who performed all the ALS manoeuvres on the scene either until the recovery of the pulse or until certified death; with very few exceptions, patients were only taken to hospital if ROSC was achieved. These characteristics, in addition to others already mentioned, differentiate this study from others published previously [4], in which ALS techniques were performed only in the hospital after being transferred, maybe having received only basic life support manoeuvres, which implies a prolongation of the time of low cerebral flow, which can affect the final results.

In our study, the analysis was performed on scene and makes it possible to have blood results even in cases of patients declared dead on scene, which was not possible in other studies analysing blood sample upon arrival at the hospital.

Thus, according to these findings, the metabolic data collected at the arrival of the team, at the beginning of ALS on scene through a venous sample [24]—undervalued in the current international guidelines [25, 26]—should be considered as a factor to be taken into account in cardiac arrest management in the out-of-hospital setting. The goal would not be just estimating the duration of the arrest and/or the quality of bystander CPR, but trying to estimate the probability of ROSC and neurological outcome, even in patients with prolonged CPR.

An interesting study with some similarities with ours revealed the association between blood variables such as pH and K+ and the prognosis of the patient [4]. However, most patients included in that study received ALS manoeuvres once in the hospital, and some confounding factors were not considered in the statistical analysis, such as the initial rhythm, presence of first respondents or the age of the patients.

The relationship between the analytical values and the clinical outcomes is remarkable, as can be seen from the magnitude of the effect of these variables: the risks of no-ROSC and no-neurological recovery increase, on average, for each unit of increase in pCO2 (mmHg), 5% and 3%, respectively. In the same way, the risks of no-ROSC and no-neurological recovery decrease, on average, for each tenth of increase in pH, by 9.4% and 9.7%, respectively. Similar values can be deduced from Table 4.

These results are coincident, in terms of the influence on these two clinical outcomes (ROSC and neurological outcome CPC I-II), with those of some of the epidemiological factors assessed in this work, such as the first monitored rhythm, the presence of first responders, the age or sex, although in the last two we only managed to establish an association with neurological recovery (Table 2).

Lactate values were analysed in previous studies for its potential prognostic value in critical-care units, with positive results as a prognostic factor to predict the patient’s progress when measured after ROSC [27]. However, in our study, with lactate measurement at the start of resuscitation, we did not observe a significant statistical relationship with ROSC or with neurological recovery. This was surprising to us. Interestingly, base excess was associated with outcomes. This may be due to the presence of other metabolites or not measured weak acids.

Figure 2 shows the parallelism between the percentages of ROSC and neurological recovery with values of pH and pCO2 closer to normal. In our study, no survival was obtained with an optimal neurological recovery in patients with a pH lower than 6.60. Of particular interest was the bimodal graph that relates kalemia with CPC I-II survival, where both the levels of hypokalemia and hyperkalemia (mainly the latter) were related to survival with worse neurological outcome, as was stated by Lin et al. [28].

Our study has some limitations. First, it was a study carried out in a single geographical area of action with specific assistance. Our good results could be explained by several factors: (1) Our EMS is based on physicians and nurses on board and in all severe cases the chief doctor and the head nurse are commissioned to collaborate with the team. (2) The scope of our service is the public area of a big city; only a scarce number of cases occur in homes, so it may probably influence the profile of patients. (3) Our response times are lower than those in other studies. (4) Training is of great importance in our institution, with frequent mandatory courses and sessions. (5) There is fluid feedback from hospitals to improve our quality standards. (6) SAMUR-PC has, as one of its main objectives, the citizenship education. The first responder programme is aimed mainly at members of security forces and firefighters.

Second, 126 cases were lost because a partial or full blood analysis could not be performed due to high or low ambient temperatures that limited the use of the EPOC® device, given that the analyser needs to be in a certain interval to be operative and reliable. The possible relationship of those extreme temperatures with the cause of the CA and the probability of recovery is unknown. The percentage of these cases is under 10% and would probably have a minor effect on statistical analysis. Anyway, it could be speculated that colder ambient conditions could improve the neurological prognosis as well as the opposite. Third, blood samples were obtained when the first venous (or intraosseous) line was achieved, but the time from the start of CPR to that achievement was not recorded. It is known that the local metabolic condition (in the limb of the venous access) may not be exactly the same of that in the rest of the body, but it is the one we can analyse and it is the best available surrogate. Previous studies used the same method. Fourth, the cause of cardiac arrest and the no-flow/low-flow time are unknown. Some aetiologies could be related to a worse initial metabolic state if the patient was already deteriorating before cardiac arrest.

Our approach is limited when trying to identify cases where CPR would be futile. This would require a different approach. It is really unlikely that we can find a threshold, except strikingly extreme values, above or below which continued resuscitation is surely futile. However, knowing which biomarkers predict outcome could be helpful when deciding to maintain a prolonged CPR effort. This could be challenging for future directions in investigation.

Finally, we have assumed monotonic relationships between continuous variables and outcomes. This is probably true for most of them but potassium levels will probably have a more complex relationship with outcome and this may underestimate the association.

Although this study was designed to find variables related with ROSC and CPC I-II survival, we feel that the importance of blood analysis on scene goes further, i.e. identifying the severity of this or other conditions so further research could follow these directions.

Conclusions

In the context of the non-traumatic OHCA, our data show the association between gas test abnormalities at the start of CPR and worse clinical outcome in terms of lower ROSC rate and higher CPC scale grade. Given the difficulties obtaining reliable data in this area, these metabolic variables at the beginning of CPR are of great diagnostic and prognostic value. Further studies should address the usefulness of these measurements at the start of resuscitation for decision-making.