Diagnostic and Prognostic Contribution of Cerebrospinal Fluid Analysis After Cardiac Arrest

Marine Paul (  mpaul@ch-versailles.fr ) ICU Hopital Andre Mignot https://orcid.org/0000-0002-0717-9555 sarah Benghanem Hopital Cochin sybille Merceron Centre Hospitalier de Versailles hugo bellut Centre Hospitalier de Versailles anne roche Centre Hospitalier de Versailles Mikhael Giabicani Centre Hospitalier de Versailles orence dumas Hopital Cochin amandine henry Centre Hospitalier de Versailles paul jaubert Hopital Cochin fabrice bruneel Centre Hospitalier de Versailles jean pierre bedos Centre Hospitalier de Versailles alain cariou Hopital Cochin stephane legriel Centre Hospitalier de Versailles

due to CA [7]. Better knowledge of such abnormalities would help to interpret CSF ndings after CA and might also assist in establishing the prognosis [5,8].
We therefore designed a retrospective study of prospectively established databases to evaluate the diagnostic and prognostic contribution of CSF ndings in patients admitted to the ICU with ROSC after CA.

Methods
We used two prospectively collected databases established at the Cochin hospital and Versailles hospital (#NCT03594318), two reference CA centres serving the southern and southwest areas of the Paris metropolis (France), respectively. Data collection was approved by the ethics committee of the French Intensive Care Society (#CESRLF_12-384 and 20-41) and the data were collected in compliance with French data protection legislation (French Data Protection Authority #MR004_2209691).

Study setting and early patient management
In France, when the emergency services receive a call reporting a suspected case of OHCA, the re department and mobile emergency unit system despatch a team to the scene. The staff in each mobile emergency unit includes at least one physician trained in emergency medicine in compliance with international guidelines [9], who performs resuscitation. Patients with in-hospital CA are initially managed by the nurses and/or bedside physician until the arrival of an emergency physician, intensivist, or anaesthesiologist, who performs resuscitation. Patients with a stable return of spontaneous circulation (ROSC) are then admitted to the intensive care unit (ICU).

Post-resuscitation diagnostic evaluation
As recommended in current guidelines [2], a standardized diagnostic workup is started immediately to allow the prompt identi cation and treatment of the cause of CA. In patients with clinical and/or electrocardiographic evidence of myocardial ischemia and in those with no obvious non-cardiac cause of CA, coronary angiography is performed at hospital arrival, before ICU admission. If prodromal symptoms or the clinical ndings suggest a respiratory or neurological cause of CA, CT of the chest or brain, respectively, may be chosen as the best rst-line investigation. When the rst-line investigation fails to detect a cause, further tests are considered [10]. Additionally, after ICU admission, laboratory tests are performed routinely to look for metabolic abnormalities or toxic substances according to the clinical history. A lumbar puncture for CSF collection is performed in patients with meningeal syndrome and when deemed appropriate by the physician in charge. All these investigations were available in both participating centres 24 h a day and 7 days a week. No post-mortem examination was performed.

Study population
All eligible patients entered into the Cochin and Versailles CA databases between January 2007 and December 2016 were included if they were older than 18 years, had stable ROSC at hospital admission, and underwent lumbar puncture as part of the aetiological CA work-up. We did not include patients who underwent lumbar puncture for other reasons or who had a traumatic lumbar puncture de ned as a CSF white cell count/red cell count <1/1000.

Study objectives
The primary objective was to assess the potential contribution of CSF analysis to the aetiological evaluation of CA. The secondary objectives were to identify factors associated with CSF abnormalities (de ned as protein >0.45 g/L and/or white cell count >5/mm 3 ) and factors associated with survival and functional outcome at ICU discharge in those patients whose CSF analysis did not contribute to the aetiological diagnosis [11].

Data collection
Demographic data and data related to the CA were collected prospectively in the two electronic databases according to the Utstein style [12]. These data included age and sex, location at CA occurrence and initial rhythm, no-ow and low-ow times, presence of a witness, bystander CPR, number of de brillations, and epinephrine use. We also recorded comorbidities, initial ECG ST-segment elevation, coronary angiography and/or CT ndings, and de nitive cause of CA. The following were collected in the ICU: use of targeted temperature management, presence of post-resuscitation shock, post-anoxic status epilepticus, and/or awakening de ned as a response to commands with a motor Glasgow Coma Scale score of 6.
To further investigate the value of CSF analysis after CA, we used standardized forms to retrospectively collect the following from the pre-hospital and ICU records: symptoms preceding CA (e.g., headache, focal signs, confusion, coma, and seizures), CSF characteristics (biochemistry, cytology, and culture results), time to CSF collection, blood sample ndings on the day of CSF collection, and CSF/serum protein quotient.
The functional outcome was assessed using the Cerebral Performance Category (CPC) at ICU discharge, and causes of death were recorded [13][14][15]. We de ned a favourable outcome as a CPC score of 1 or 2 at ICU discharge.

Statistical analysis
Quantitative parameters were described as median (interquartile range [IQR]) and qualitative parameters as number (percentage). We compared categorical variables using Fisher's exact test and continuous variables using the Wilcoxon rank-sum test.
We rst tested univariate associations between CA characteristics and whether CSF analysis contributed to the aetiological diagnosis of CA. We then looked for associations linking CA features to speci c CSF abnormalities and to survival and functional outcome at ICU discharge.
All tests were two-sided and p values <0.05 were considered signi cant. Analyses were performed using R statistical software version 3.1.2 (R Foundation for Statistical Computing, Vienna, Austria, http://www.Rproject.org. accessed September 12, 2019). Figure 1 is the patient ow chart. Of the 1984 patients admitted with stable ROSC after CA, 65 (3.3%) had a lumbar puncture and were included in the study.

Results
Characteristics and diagnostic workup Table 1 and Table S1 report the patient characteristics and the diagnostic investigations performed to identify the cause of CA. Figure 2 shows the rst-line, second-line, and third-line investigations. Overall, cerebral CT was done in 52 (80%) patients, cerebral MRI in 5 (8%) patients, coronary angiography in 31 (48%) patients, and chest CT in 33 (51%) patients. The cause of CA was identi ed in 52 (80%) of the 65 patients and was respiratory in 18 (27%) patients, neurologic in 16 (25%) patients, cardiac in 7 (11%) patients, metabolic in 7 (11%) patients, and septic in 4 (6%) patients.
Contribution of cerebrospinal uid analysis to the aetiologic diagnosis LP were assessed in cases of suspicion of a neurological cause for cardiac arrestin 69% of cases. CSF analysis identi ed a neurological cause of CA in 6/65 (9%) patients, including 3 with subarachnoid haemorrhage, 2 with non-speci c encephalitis, and 1 with bacterial meningitis. In 4 of these 6 patients, the lumbar puncture was performed post-mortem. Table 1 reports the results of the univariate analysis of factors associated with the CSF analysis contributing to the aetiological diagnosis of CA. In the patients with a contributory CSF analysis, lumbar puncture was mostly the second-line investigation, with a median time of 1.5 days (IQR, 1-2) after ICU admission. Neurologic prodromal symptoms before CA were more common in the patients whose CSF analysis was contributory compared to the other patients (83% vs. 68%), although the difference was not signi cant (p=0.66).

Patients with nonspeci c cerebrospinal uid abnormalities
We excluded the 6 patients whose CSF analysis was contributory (including 4 with post-mortem lumbar puncture) and the additional 6 patients with post-mortem lumbar puncture. Of the remaining 53 patients, 37 (70%) had abnormal CSF ndings, which are reported in Table 2. Table 3

Patient outcomes
Overall ICU mortality was 70% (46/65). Table 1 shows the causes of death. All 6 patients whose CSF analysis contributed to the aetiological diagnosis died. Nonspeci c CSF abnormalities were more common in patients with poor outcomes de ned as CPC 3, 4, or 5 (73% versus 27% of those with CPC 1 or 2), although the difference was not statistically signi cant (p=0.06). Table S2 reports the CSF features in patients with favourable versus unfavourable outcomes. The only factor signi cantly associated with outcome was the CSF/serum protein quotient (p=0.017), with higher values in patients with worse outcomes.

Discussion
To our knowledge, this study provides the rst detailed information on the diagnostic and prognostic contribution of CSF analysis to the aetiological diagnosis of CA. Only 3.3% of all patients admitted to the ICU with stable ROSC after CA underwent CSF analysis, which contributed to the aetiological diagnosis in 6 (9.2%) patients, although in 4 this contribution was obtained only post-mortem. Of the patients alive at the time of lumbar puncture, many (69.8%) had nonspeci c CSF abnormalities, among which the CSF/serum protein quotient was signi cantly associated with the outcome.
Our study design provides a pragmatic view of the contribution of CSF analysis to the aetiological diagnosis of CA in patients with sustained ROSC at hospital admission. Lumbar puncture was performed only very rarely in our study. Few previously published data are available with which to compare our results. Most studies of CSF analysis after CA focused on the neuroprognostication accuracy of CSF biomarkers re ecting neuronal damage [5,6,8]. We are not aware of previous studies investigating CSF analysis for the aetiological diagnosis or the presence of CSF abnormalities unrelated to the aetiology. In previous studies, CSF analysis was performed in 5.3% of patients with neurological causes of CA and stable ROSC at hospital admission, chie y as part of the aetiological workup [4], and in 40% of patients with CA complicating convulsive status epilepticus [4,16].
Given, the low incidence of noncardiac causes of CA, recent guidelines focus on the indications of coronary angiography, cerebral CT, and chest CT. Important factors are the patient's medical history; the presence of cardiac, respiratory, or neurologic prodromal symptoms, the circumstances of CA onset, and the physical ndings on the scene. In practice, lumbar puncture is not a rst-line investigation, unless there is evidence of a neurological cause whose identi cation may be helped by CSF analysis. Obstacles to lumbar puncture include anticoagulant and/or antiplatelet treatments, and concern about inducing cerebral herniation. Thus, cerebral CT may be required before lumbar puncture is performed. As expected, lumbar puncture was mainly performed as a second-or third-line investigation in our study, predominantly in patients with neurological prodromal symptoms before CA. Interestingly, 10 (15%) lumbar punctures were post-mortem and contributed to the diagnosis in 4 patients. This nding suggests that the situations in which CSF analysis may be helpful may not be recognised su ciently early. Work is clearly needed to determine the indications of lumbar puncture after CA. An optimal aetiological workup is crucial to determine when speci c aetiological treatments are appropriate, thus improving patient outcomes. In previous studies, ICU survival was higher when the aetiology was identi ed [3,17]. In addition, identifying the cause may allow measures to minimise the risk of recurrent CA. Finally, knowledge of the causes of CA is important from a public health perspective. Lumbar puncture identi ed the cause of CA in 9% of our patients, although this proportion dropped to 3% when only patients alive at the time of lumbar puncture were considered.
Over two-thirds of our patients without neurological causes of CA had nonspeci c CSF abnormalities, of which the most common was an increase in protein (73%), followed by an increase in white cells (27%). Several hypotheses can be raised to explain these ndings. First, we retrospectively identi ed neurological prodromal symptoms in 28 of the 59 patients whose CSF analysis did not contribute to the aetiological diagnosis, and many of these patients did not undergo a comprehensive neurological workup. For instance, cerebral MRI was performed in only 5 of these patients. Moreover, new tools for diagnosing auto-immune and/or infectious encephalitis were not available during the study recruitment period [18,19]. Thus, some of the patients whose CSF abnormalities were considered nonspeci c may have had undiagnosed neurological conditions. Another hypothesis is that BBB disruption after CA may result in CSF abnormalities. In healthy individuals, most of the proteins found in the CSF are derived from the serum, although some are synthesized by the choroid plexus or within the brain. The passage of serum protein into the CSF varies with the condition of the BBB [20,21]. Normal BBB permeability is de ned as a CSF/serum albumin quotient <0.007 [22,23]. BBB disruption may allow the passage of greater amounts of protein from the serum to the CSF. CSF ndings may be di cult to interpret in patients with brain injury, as reported in a study of status epilepticus [24].
We identi ed post-resuscitation shock as factor associated with having nonspeci c CSF abnormalities. The systemic in ammation seen in post-resuscitation shock may cause BBB alterations, as described in acute sepsis and cirrhosis, [25,26]. Moreover, patients presenting with confusion to coma before CA and who demonstrated oedema on cerebral CT Scan were more likely to have nonspeci c CSF abnormalities. In the setting of primary brain injury, brain in ammation could cause BBB alteration as described in stroke and status epilepticus [24,27,28].
CSF changes may also occur in response to anoxic neuronal damage. Thus, elevated levels of proin ammatory cytokines in CSF have been reported after CA [29,30]. HMGB1 (high-mobility group box 1), released or secreted by necrotic brain cells, may act as an early in ammation trigger inducing the local recruitment of pro-in ammatory cytokines, independently of BBB alterations. [6] An increase in the levels of neuronal speci c enolase, protein S100B, T-tau protein, neuro lament were also reported [6,31].Finally, CSF abnormalities can be induced by many factors including drugs, spinal cord compression, diabetes, and polyradiculoneuritis [32]. In uence of systemic and neuro in ammation after CA on CSF protein level could not be further explored because of the non-availability of albumin CSF/blood ratio or speci c MRI exploration to assess the BBB function [33,34].
ICU mortality was 70% in the overall population of patients with lumbar puncture after CA. Of the 6 patients whose CSF analysis contributed to the diagnosis, 2 had the lumbar puncture done while alive but died subsequently and 4 had the lumbar puncture done post-mortem. Identifying a neurological cause of CA has been reported to carry a very poor prognosis [4,35]. In our study, ICU mortality in the patients whose CSF analysis did not contribute to the diagnosis but showed nonspeci c abnormalities was 73%. A higher CSF/serum protein quotient was the only variable signi cantly associated with a poor outcome. Similarly, a prospective study in 21 patients found that the CSF/serum albumin quotient was higher in the subgroup of 10 patients with poor outcomes than in the other patients [36]. These ndings support the existence of BBB disruption after CA. Finally, in our cohort, 56% of deaths in case of nonspeci c CSF abnormalities were ascribed to withdrawal of life-sustaining treatments due to severe post-anoxic encephalopathy.
Our study has several limitations. First, given the retrospective nature of this study design and our sample size, the extent to which our ndings apply to the full spectrum of patients with CA is unclear. We included consecutive patients with lumbar puncture after ICU admission with stable ROSC after CA, but lumbar puncture was not performed according to prede ned criteria, either in the ICU or post-mortem. Moreover, the two participating ICUs were in high-volume centres, and their recruitment may not re ect that of ICUs overall. However, one of the centres was a referring university hospital and the other a tertiary referral hospital. Second, we considered only CSF analysis performed at the early phase after CA, as part of the emergent aetiological workup. Delayed CSF analysis may provide important information. One study found that the CSF/serum albumin quotient increased between 24 h and 72 h after ROCS, and others reported an increase in protein levels after 2-3 weeks [7]. However, our focus was on the potential usefulness of CSF analysis for the aetiological diagnosis and the prognosis. Finally, CSF albumin values were not available, and we did not adjust the CSF protein values on age [37].

Conclusion
In conclusion, although rarely performed after CA, lumbar puncture may contribute to the diagnosis of a neurological cause. In our study, CSF analysis as a second-line investigation identi ed a neurological cause in 9% of patients. Nonspeci c CSF abnormalities are common after CA, perhaps due to BBB disruption, and may have prognostic signi cance. Further studies are warranted to further assess these hypotheses. Availability of data and material The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Con ict of interest statement
The authors declare that they have no con ict of interest.

Funding
There was no funding for the development and writing of this commentary.
Authors' contributions MP and SL wrote the rst draft of the paper. All authors approved the nal version of the manuscript.   ICU, intensive care unit; LP: lumbar puncture; ROSC: return of spontaneous circulation. † Abnormal CSF was de ned as CSF white-cell count >4/mm 3 and/or CSF protein >0.45 g/L. Figure 1 Patient ow diagram ROSC denotes return of spontaneous circulation and CA cardiac arrest.

Figure 2
Diagnostic work-up in 65 patients with a lumbar puncture after cardiac arrest CT denotes computed tomography.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. ESM0308.docx