Placing intracerebral probes to optimise detection of delayed cerebral ischemia and allow for the prediction of patient outcome in aneurysmal subarachnoid haemorrhage

Abstract

Cerebral microdialysis could be useful to detect delayed cerebral ischemia in aneurysmal subarachnoid haemorrhage patients. The optimal location of the probes, however, remains controversial. Here, we determined the vascular territories with the highest infarct risk in relation to aneurysm location to define probe implantation guidelines. These guidelines were retrospectively validated by studying the likelihood of probe to fall in a secondary infarct area, and by analysing their influence to predict patient outcome. The vascular territories with highest risk of infarction were the anterior cerebral arteries for anterior communicating artery aneurysms and the ipsilateral middle cerebral artery for internal carotid artery, posterior communicating artery and middle cerebral artery aneurysms. When cerebral microdialysis probes had been implanted in these territories, 79% were located within an infarcted area versus 54% when they were implanted in other territories. Delayed cerebral ischemia was detected only when the probe was located within a brain area later affected by secondary infarction, which could justify the use of implantation guidelines. Moreover, individual patient outcomes could be predicted when probes were placed in the brain territories as suggested by this study. Thus, a precise probe placement algorithm can improve delayed cerebral ischemia detection sensitivity and allow for a better prediction concerning patient outcome.

Introduction

Delayed cerebral ischemia (DCI) is a devastating complication of aneurysmal subarachnoid haemorrhage (ASAH) that occurs in about 30% of patients surviving the ictus.¹ The pathophysiology of DCI is not yet fully known and involves multiple factors, which may act in concert to produce cerebral damage. Among the causative factors of DCI, it is possible to include cerebral vasospasm, spreading depolarization and spreading ischemia,^2,3 microvascular thrombosis,^4,5 seizures,⁶ impaired autoregulation,⁷ sustained microvascular spasm⁸ and inflammatory pathways.^9,10 Whatever its biological origin(s), DCI is a major cause of cerebral infarction and results in death or severe disability in about half of affected patients.^11,12

Clinical diagnosis of DCI is difficult, especially in unresponsive comatose patients. In these cases, diagnosis requires detection of brain ischemia by computed tomography (CT) or magnetic resonance imaging (MRI), preferably with identification of cerebral vasospasm by trans-cranial Doppler or cerebral angiography.¹³ However, MRI and angiography cannot be repeated over short intervals. Cerebral microdialysis (CMD) or brain tissue oxygen tension (PbtO₂) probes implanted directly in brain parenchyma, however, do allow continuous monitoring and may detect early pathological changes predictive of DCI.^13–18 Neuro-monitoring of brain metabolism has been associated with long-term patient outcome^19,20 and even though the possibility of individual prognosis based on these data has not been demonstrated to date, such information can inform delicate clinical decisions such as haemodynamic/volemic optimization or local vascular interventions in case of significant cerebral vasospasm.

It is reasonable to assume that the sensitivity of this type of local neuro-monitoring would be linked to the location of probe placement within brain area later affected by the secondary infarction. However, the true impact of probe placement within an infarct zone on detection sensitivity is not well documented. Moreover, it is possible that with optimal probe placement, individualized prediction of long-term patient outcome could be improved. Precise guidelines for optimal probe placement enabling the highest probability of monitoring an infarct zone in DCI are therefore needed. The consensual guidelines available to date were produced at three meetings that took place in 2004, in 2014 and 2015.^21–23 The first guidelines recommended that the cerebral probe be implanted ‘in the parent vessel territory’.²¹ Such a recommendation is imprecise, however, especially for internal carotid artery (ICA) or posterior communicating artery (PCoA) aneurysms, which supply virtually a whole hemisphere, or for aneurysms affecting posterior arteries that supply blood to occipital regions, the brainstem and cerebellum.

The second consensus states that ‘the location of the microdialysis probe depends on the diagnosis, the type and location of brain lesions, and technical feasibility’.²² Finally, the third consensus recommended the placement of the catheter ‘in the watershed anterior cerebral artery–middle cerebral artery territory (frontal lobe) on the same side as the maximal blood load seen on CT or the ruptured aneurysm’.²³ Both recommendations do not provide exact probe positioning guidelines corresponding to aneurysm location. Recently, research teams from Bern and Frankfurt have suggested a simple algorithm (Bern–Frankfurt algorithm), wherein the probe should be placed in the ipsilateral anterior cerebral artery (ACA) territory for ACA aneurysms, in the right ACA territory for anterior communicating artery (ACoA) aneurysms, and in the ipsilateral middle cerebral artery (MCA) territory for MCA, ICA and all posterior aneurysms.²⁴

In this study, we determined the vascular territories with the highest risk of secondary infarction in a cohort of 321 ASAH patients enrolled in the Lyon Neurological Hospital. We hypothesized that using implantation guidelines to place the probes in these territories would provide the highest probability of monitoring an infarcted zone in case of DCI, and therefore, more reliable information to guide treatment. We confirmed the efficacy of these guidelines on probe placement by analysing location retrospectively in patients who had received CMD/PbtO₂ monitoring and determined the rate at which the probe was placed within an infarcted area. We validated the need to maximize the probability of placing the probe into brain area with a high risk of secondary infarction by comparing the results of CMD/PbtO₂ monitoring according to the congruence of the probe placement with the area of infarction. Finally, we examined the influence that proper probe placement had on predicting long-term patient outcome based on CMD/PbtO₂ monitoring.

Material and methods

Study populations and inclusion criteria

The study was conducted in accordance with the Helsinki Protocol, and approved by the ethics committee of the Lyon University Medical Center (French clinical research program HCL99.173) and by institutional review board of Lyon (‘Comité de Protection des Personnes Sud-est IV’, ref: L12-135). Considering the pathology, all conscious patients and legal representative of all comatose patients gave written informed consent. Data were collected from the database of the neurological intensive care unit (ICU) at Lyon Neurological Hospital (Hospices Civils de Lyon). The total population was separated in two different populations that were retrospectively analysed in this study: (a) 332 ASAH to identify the most at-risk vascular territories for secondary infarction based on the location of the aneurysm, (b) 61 ASAH patients who had received CMD/PbtO₂ monitoring to validate the influence of this new implantation algorithm by comparing the congruence of the probe placement with secondary infarction when probes were implanted following versus not following our algorithm, to justify the interest of probe placement guidelines by analysing neurochemical data and to examine the influence of these implantation guidelines for predicting long-term patient outcome based on CMD/PbtO₂ monitoring.

To identify a new probe placement algorithm, we retrospectively analysed the charts of all ASAH patients admitted to the neurological ICU (N = 207, including 7 from the CMD/PbtO₂ population) between January 2011 and April 2014. In an attempt to increase the number of secondary infarction cases, we decided to also include 125 severe ASAH patients, in which multi-modal monitoring had been employed (including 54 from CMD/PbtO₂ population), hospitalized between 2000 and 2010 in the neurological ICU. Aneurysm location, infarcted territory and demographic data were recorded for these 332 patients.

Among the 332 patients included in the first part of the study on infarct location, a subset of 61 patients who had received CMD/PbtO₂ monitoring was included in the second part of the study assessing neurochemical monitoring efficiency. All patients who underwent CMD/PbtO₂ monitoring met the following criteria: severe ASAH, as indicated by score of 4 or 5 in World Federation of Neurologic Surgeons (WFNS) scale at admission,²⁵ absence of return to consciousness or contraindication to stop sedation (i.e. intracranial hypertension, cardiac or respiratory complications). A PbtO₂ probe (LICOX, Integra NeuroSciences, Saint-Priest, France) and a CMD catheter (CMA 70; M Dialysis AB, Johanneshov, Sweden; cut-off, 20 kDa; flow rate, 0.3 µL/min) were inserted via a triple-lumen catheter and placed as recommended by the 2004 consensus,²¹ except in the following circumstances: (a) aneurysm of the posterior circulation and (b) logistical limitations, such as a decompressive craniectomy, extraventricular derivation or brain lobectomy precluding access to recommended area. Conventionally, both probes were implanted in the ICU on the third day after bleeding. CMD monitoring was observational only. Therapeutic interventions were implemented only when PbtO₂ was < 15 mmHg in spite of maximal inspired oxygen content and/or cerebral perfusion pressure optimization.

Only patients who met the following criteria were finally included: (a) age ≥ 18 years; (b) ASAH confirmed by CT and cerebral angiography with precise identification of the haemorrhagic aneurysm; (c) admitted ≤ 72 h after ASAH; (d) aneurysm treated within 24 h of hospitalization and (e) availability of all brain imaging data. All patients were treated in the neurological ICU according to standard ASAH management guidelines.^26,27

Identification of secondary infarction from DCI

In all patients, CT scans were obtained routinely (a) at admission, (b) after aneurysm treatment, (c) following brain monitoring or external ventricular drain (EVD) placement, (d) after any neurological change or unexplained, sustained intracranial pressure elevation and (e) at distance from the haemorrhagic event to evaluate the brain injury. If all practical conditions were fulfilled (availability of the equipment, possibility of transporting the patient, absence of invasive monitoring, etc.), MRI was preferred when a suspicion of DCI existed (neurological deterioration, increase in velocities evidenced by transcranial Doppler, etc.) or to evaluate brain injuries at distance from the haemorrhagic event. We should note that access to MRI in our clinical centre was difficult for neurological intensive care patients in the years before 2006, and the use was increasing thereafter. In accordance with recommendations, infarction from DCI was identified as cerebral infarction on a CT or MRI scan, not present on the CT or MRI scan between 24 and 48 h after early aneurysm occlusion, not attributable to other causes such as surgical clipping or endovascular treatment and detected within 6 weeks post SAH, or present on the last CT or MRI scans made before death within 6 weeks.¹² All the CT and/or MRI scan were retrospectively reviewed by an experienced radiologist who was blinded to the monitoring data and who determined the affected vascular territories if infarction was present. Finally, the presence of infarction secondary to DCI was agreed on by consensus between the authors.

Identification of a new probe implantation algorithm

Three hundred thirty-two ASAH patients were retrospectively analysed to identify the vascular territories with highest risk of infarction depending on the aneurysm location. The patients were classified according to haemorrhagic aneurysm location and the occurrence of cerebral infarction based on reviews of their neuroimaging charts. For each kind of aneurysm, the frequency of a specific infarcted territory was determined. An implantation algorithm related to aneurysm location was defined to maximize the probability of monitoring brain areas with a high risk of infarction.

Retrospective validation of this new implantation algorithm

To validate the interest of this new implantation algorithm, we then estimated the probability of probe placement being in an area of secondary infarction when this algorithm was used in a group of 61 CMD-monitored patients. For all patients in whom infarction was observed, imaging data were reviewed to analyse the association between the probe implantation site and the vascular territory of the cerebral infarct (Figure 1). The percentage of cases in which the probe was located in an infarcted brain area was computed for each aneurysm type.

Figure 1.

Example cerebral microdialysis (CMD) probe locations relative to infarction. Representative non-contrast computed tomography images of (a) CMD probe catheter within an infarcted territory (ruptured left middle cerebral artery aneurysm) and (b) CMD probe catheter outside of an infarcted territory (anterior communicating artery aneurysm). Red circle: microdialysis catheter; white arrow: location of secondary infarct.

Justification of the need to find probe placement guidelines by analysing neurochemical parameters when the probe was placed in brain area later affected by secondary infarction or not

It is reasonable to assume that the sensitivity of local monitoring would be linked to the accuracy of the probe placement within a brain area later affected by secondary infarction. However, the true impact of probe placement within an infarct zone on detection sensitivity is not well documented. So we decided to compare the neurochemical changes in patients who had received CMD/PbtO₂ monitoring (the same 61 patients) by taking into account the relation between the position of the probe and the location of the secondary infarcts. Thus, these patients were retrospectively studied and were divided into three groups: (a) no detectable secondary infarction, (b) secondary infarct apparent in imaging scans with CMD probe outside infarct area and (c) secondary infarct with CMD probe within infarct area. To analyse the interest of the congruence of the probe placement with secondary infarction, we decided to analyse variations of lactate/pyruvate (L/P) ratio, as recommended recently,²² in these three groups of patients. We first compared group median L/P ratio values for the last 12 h of monitoring and secondarily we compared the numbers of hours (consecutive or not) of L/P ratio > 40 before the first scan showing infarction, or over the whole neuromonitoring period for the patient without secondary infarction, between the three groups.

Impact of the probe placement guidelines on outcome prediction with CMD/PbtO₂ monitoring

We examined the influence of these implantation guidelines for predicting long-term patient outcome based on CMD/PbtO₂ monitoring. Because metabolic data obtained after a secondary infarction would be too late to enable intervention, we analysed neuro-monitoring data gathered before the first scan showing infarction. Patients were divided into two outcome groups according to their Glasgow outcome scale (GOS) score²⁸ at 6 months: unfavourable (GOS 1–3) and favourable (GOS 4–5). We conducted three consecutive receiver operating characteristic (ROC) curve analyses to determine the predictive performance of using the numbers of hours spent with CMD L/P ratio surpassing the pathological value of 40.^29,30 We performed this: (a) for all patients, (b) for patients with a probe implanted in accordance with the algorithm described above (evaluation of the benefit of the new algorithm) and (c) for patients with a probe correctly placed in a secondary infarct area or without infarction (efficiency of patient prognosis in an idealized situation).

Statistical analysis

Clinical and demographic (age and onset delay of infarction) data were compared by analysis of variance (ANOVA) for normal distributions or by Mann–Whitney tests for non-normal distributions. Chi-square tests were used for the other clinical/demographic parameters (female/male ratio, WFNS,²⁵ or GOS score,²⁸ type of aneurysm treatment, proportion of secondary infarction). The percentages of patients with a probe implanted in subsequent infarction territory were compared with a Chi-square test. Kruskal–Wallis tests (followed by Mann–Whitney multiple comparison tests) were used to compare group L/P values. Since we performed two consecutive comparisons between the three groups, the alpha level was decreased from p = 0.05 to p = 0.025 (according to Bonferroni’s correction). For the ROC curve analysis, only parameters that reached an area under the curve (AUC) > 0.8 were considered significant for predicting long-term outcome. Parametric data are expressed as means ± standard deviations; non-parametric data are expressed as medians (with 25th–75th percentile ranges). Differences were considered statistically significant at p < 0.05. The statistical analysis was performed with MedCalc version 11.3.1.0 (http://www.medcalc.be).

Results

Identification of probe placement guidelines for monitoring brain areas with high risk of infarction

Three hundred thirty-two ASAH patients were retrospectively analysed to identify the vascular territories with highest risk of infarction depending on the aneurysm location and to deduce probe placement guidelines.

Demographic data

Of 332 patients, 11 were excluded: 4 due to incomplete imaging data, 3 due to imprecise identification of the haemorrhagic aneurysm and 4 because admission was > 72 h after haemorrhagic stroke or the aneurysm was treated more than 24 h after hospitalization. The study population demographics are summarized in Table 1.

Table 1.

Characteristics of patients who were analysed to identify the vascular territories with the highest infarct risk in relation to aneurysm location.

Parameters	Total N	ACoA	MCA	PCoA	ICA	BA	VA	PICA	ACA	PCA	p
Number of cases	321	121	73	42	34	19	7	14	8	3	–
Sex (F/M)	189/132	66/55	40/33	31/11	23/11	10/9	3/4	9/5	4/4	3/0	0.34
Age, mean ± SD	52 ± 14	55 ± 14	48 ± 13	51 ± 14	52 ± 14	49 ± 11	49 ± 13	52 ± 11	57 ± 13	64 ± 12	0.02
WFNS score											0.52
Grade 1–3	133	53	32	16	17	5	2	5	2	1
Grade 4–5	188	68	41	26	17	14	5	9	6	2
Fisher score											0.67
Grade 1	18	3	4	3	1	3	1	1	1	1
Grade 2	26	12	5	5	4	0	0	0	0	0
Grade 3	27	8	9	4	2	2	0	1	1	0
Grade 4	250	98	55	30	27	14	6	12	6	2
Aneurysm treatment Coiling/clipping/others	260/50/11	115/4/2	34/36/3	39/2/1	25/6/3	18/0/1	7/0/0	11/2/1	8/0/0	3/0/0	<0.0001
Secondary cerebral infarct
Incidence (%)	38.9	37.2	38.4	47.6	38.2	47.4	28.6	42.9	25	0.0	0.86
Onset median (h) (25th–75th percentile)	192	210	191	203	165	144	334	118	229	–	0.53
	(116–264)	(116–269)	(122–261)	(106–303)	(141–194)	(98–288)	(312–355)	(102–158)	(204–253)
GOS score at 6 months											0.32
GOS 1–3 (%)	63.1	67.8	62.7	53.0	66.7	73.3	57.2	36.4	57.2	50
GOS 4–5 (%)	36.9	32.2	37.3	47.0	33.3	26.7	42.8	63.6	42.8	50

p < 0.05 was considered significant (indicated in bold).

ACA: anterior cerebral artery; ACoA: anterior communicating artery; BA: basilar artery; F: female; GOS: Glasgow outcome scale; ICA: internal carotid artery; M: male; MCA: middle cerebral artery; PCA: posterior cerebral artery; PCoA: posterior communicating artery; PICA: posterior inferior cerebellar artery; SD: standard deviation; VA: vertebral artery; WFNS: world federation of neurosurgical societies.

Detection of secondary cerebral infarction

Among the 321 patients included in this study, 181 were followed by serial CT scans, 140 with CT scan and at least one diffusion-weighted MRI. MRI was performed increasingly in the 2000–2014 period, most often for assessment of brain injuries at distance (N = 114), and when DCI was suspected (N = 26). Secondary cerebral infarcts were identified in 125 patients, and the first imaging test showing this delayed infarct was a CT scan in 86.4% of cases (N = 108) and a diffusion-weighted MRI in 13.6% of cases (N = 17). Among the 108 patients with delayed infarcts first identified by CT scan, 30 patients had a diffusion-weighted MRI at distance (post-bleeding median delay: 23 (16–57) days), which always confirmed the presence of delayed infarcts. In the 17 patients with delayed infarcts first detected using MRI, 14 patients had subsequent serial CT scan. In 11 of these 14 patients, infarction was also observed on the CT-scan a posteriori (agreement of 78.6%), whereas in three patients, CT scan could not detect the infarct initially observed using MRI. This was consistent with the better sensitivity of MRI compared with CT scan to detect infarcts. Overall, relying on CT for more than half of the patients probably underestimated the frequency of DCI by about 15 patients. On the other hand, the small number of MRIs used to detect secondary infarcts in our study allowed us to detect slightly more cases than if we had used only CT scan (three to four patients). Therefore, including patients with imaging data obtained with MRI did not fundamentally change the detection of DCI compared with including only patients with CT scan.

Distribution of secondary infarcts

Infarction occurred in 37.2% (N = 45/121) of patients with a ruptured ACoA aneurysm. The ACA territories were affected in 89% of these cases (40/45), with 8/45 (18%) only in the right territory, 8/45 (18%) only in the left territory and 24/45 (53%) in both territories (Figure 2(a)).

Figure 2.

Topographical distribution of secondary infarcts in aneurysmal subarachnoid haemorrhage patients. Distribution of secondary infarcts in relation to location of associated ruptured aneurysm on the (a) anterior communicating artery (ACoA, N = 45), (b) middle cerebral artery (MCA, N = 28), (c) internal carotid artery (ICA, n = 13) and (d) posterior communicating artery (PCoA, n = 20). ACA: anterior cerebral artery.

Infarction occurred in 38.4% (N = 28/73), 38.2% (N = 13/34) and 47.6% (N = 20/42) of ruptured MCA, ICA and PCoA aneurysms, respectively (Figure 2(b) to (d)). The most affected area for infarction was the ipsilateral MCA territory (24/28, 85.7% for MCA; 10/13, 76.9% for ICA and 15/20, 75% for PCoA).

Cerebral infarction was observed in 42.9% of posterior inferior cerebellar artery (PICA) aneurysms (6/14). In 83.3% of these patients (5/6), the infarction affected the ipsilateral PICA territory. The ipsilateral MCA or ACA territories were infarcted in 2/6 ruptured PICA aneurysm cases (33%).

Infarction occurred in 42.3% of vertebro-basilar vessel aneurysm cases (9/19 for basilar artery (BA), 2/7 for vertebral artery (VA)). In these 11 cases, the most frequently affected infarct area was the vertebro-basilar circulation territory (81.8%, 9/11). Other territories were also affected in 6/11 patients, including an ACA territory (45.5%, 5/11), the left MCA territory (54.5%, 6/11) and the right MCA territory (27.2%, 3/11).

Implantation guidelines for monitoring brain areas with high risk of infarction

Analysis of infarcted-territory frequencies yielded a clear pattern, allowing for optimizing monitoring location (Table 2). In ACoA aneurysms, infarction affected an ACA territory with a 71% probability, and for MCA, ICA and PCoA aneurysms, infarction affected the ipsilateral MCA territory with a probability in the range of 75–86% (Table 2). With aneurysms in the posterior circulation (PICA, BA, VA), the parent vessel territory was more frequently affected by infarction, an area poorly accessible for probe implantation. Further, if probes were to be implanted in left MCA or ACA territories in these cases, there would be a less than 50% probability that implantation was in area of infarct (47% (8/17) for the left MCA territory; 41% (7/17) for the left ACA territory).

Table 2.

Comparison of 2004 consensus guidelines, Bern/Frankfurt algorithm guidelines and predictive data, and our algorithm guidelines and predictive data.

Aneurysm location	2004 consensus meeting (12)	Bern/Frankfurt algorithm (15)	Present study
ACoA	Left or right ACA	Right ACA (78%)	Left or right ACA (71%, 83%)
MCA	Ipsilateral MCA	Ipsilateral MCA (89%)	Ipsilateral MCA (86%, 63%)
ICA	Ipsilateral hemisphere	Ipsilateral MCA (95%)	Ipsilateral MCA (77%, 100%)
PCoA	Ipsilateral hemisphere	Ipsilateral MCA (ND)	Ipsilateral MCA (75%, 75%)
PICA	Ipsilateral PICA	– (ND)	Ipsilateral MCA or no implantation (33%, ND)
BA/VA	posterior territories	Right MCA (23%)	MCA or no implantation (27–54%, ND)

Successful monitoring rates are in parenthesis. Bold: percent probability of infarct in vessel territory extracted from the study with the 321 patients; italic: percent probability of probe within infarct zone when following the guidelines extracted from the retrospective analysis of the 61 patients with local monitoring.

ACA: anterior cerebral artery; ACoA: anterior communicating artery; BA: basilar artery; ICA: internal carotid artery; MCA: middle cerebral artery; ND: not determined; PCoA: posterior communicating artery; PICA: posterior inferior cerebellar artery; VA: vertebral artery.

Retrospective validation of implantation guidelines and justification of the need to define implantation guidelines

We retrospectively analysed neuroimaging charts and neurochemical data of 61 ASAH patients who underwent CMD/PbtO₂ monitoring to (a) validate the influence of this new implantation algorithm by comparing the congruence of the probe placement with secondary infarction when probes were implanted following versus not following our algorithm, and to (b) justify the interest of probe placement guidelines by analysing neurochemical data.

Demographic data

This population is a subset of the 321 patients previously included to determine the probe implantation guidelines. It included 30 ACoA, 16 MCA, 10 PCoA, 5 ICA, aneurysm cases (total N = 61). Secondary infarction occurred in 45 of the 61 analysed cases (73.8%). The distribution of secondary infarcts in the subset of patients receiving intracerebral probes as a function of the location of the aneurysm was not statistically different from the rest of the population (χ² test, ACoA: p = 0.42, MCA/ICA/PCoA: p = 0.52). Secondary cerebral infarcts were first identified by CT in 88.8% of patients (N = 40) and by diffusion-weighted MRI in 11.2% (N = 5) that was not different from the rest of population (p = 0.54). However, 17.5% of the infarcts identified first in CT scans were then confirmed by diffusion-weighted MRI.

At 6 months, 45 patients had an unfavourable outcome (GOS 1–3, 73.8%) and 16 had a favourable outcome (GOS 4–5, 26.2%). Concerning demographic characteristics (Table 3), the aneurysm patient groups were homogeneous in terms of sex ratio, age, distribution of WFNS or Fisher grade, incidence of secondary infarction except for the type of aneurysm treatment as clipping was the type of treatment used in most MCA aneurysms while the coiling was more likely in other types of aneurysms.

Table 3.

Characteristics of patients who underwent CMD/PbtO₂ monitoring as a function of aneurysm location.

Parameters	ACoA	MCA	PCoA	ICA	p
Number of cases	30	16	10	5	–
Sex (F/M)	15/15	7/9	9/1	2/3	0.09
Age, mean ± SD	49 ± 11	44 ± 7	48 ± 9	52 ± 7	0.17
WFNS score
Grade 1–3	4	1	0	0	0.48
Grade 4–5	26	15	10	5
Fisher score
Grade 1	0	0	0	0	0.60
Grade 2	1	0	1	1
Grade 3	3	1	1	0
Grade 4	26	15	8	4
Aneurysm treatment Coiling/clipping	30/0	4/12	9/1	5/0	<0.0001
Secondary cerebral infarct
Incidence (%)	73.3	62.5	80	100	0.39
Onset median (h) (25th–75th percentile)	231	176	203	165	0.36
	(140–404)	(152–245)	(98–266)	(132–173)
GOS score at 6 months
GOS 1–3 (%)	80	62.5	60	100	0.22
GOS 4–5 (%)	20	37.5	40	0

p < 0.05 was considered significant (indicated in bold).

ACA: anterior cerebral artery; ACoA: anterior communicating artery; BA: basilar artery; F: female; GOS: Glasgow outcome scale; ICA: internal carotid artery; M: male; MCA: middle cerebral artery; PCA: posterior cerebral artery; PCoA: posterior communicating artery; PICA: posterior inferior cerebellar artery; SD: standard deviation; VA: vertebral artery; WFNS: world federation of neurosurgical societies.

Overall the population of patients who received CMD and/or PbtO₂ probes had more severe WFNS scores (WFNS 1–3: N = 5 vs. N = 128; WFNS 4–5: N = 56 vs. N = 132, respectively; p < 0.0001), and more severe Fischer scores (Fisher 1–2: N = 3 vs. N = 41; Fisher 2–4: N = 58 vs. N = 219, respectively; p = 0.03) than the rest of the patients. They also presented cerebral hematoma more often (65.6% vs. 42.6%, p = 0.002). These differences could explain the increased likelihood of delayed infarcts in these patients compared with the rest of the population included in the study.

Retrospective analysis of CMD probe implantation

The likelihood of monitoring in an infarcted territory (Figure 1) was determined by retrospective analysis of the 61 patients who received CMD/PbtO₂ monitoring, which were demographically described previously.

Probes were implanted in the territories that would have been indicated by our algorithm in 46 patients, and implanted in other brain areas in 15 patients. Notably, the probe was implanted in other brain areas due to the presence of a craniectomy ipsilateral to the aneurysm (N = 6), lobectomy (N = 1) or EVD placement that obstructed the probe placement recommended in the guidelines (N = 8). Overall, when probe position was consistent with the proposed algorithm as discussed above, the CMD probe was finally identified within a secondary infarct more frequently (79.4% of cases) than when it was placed outside the location recommended by our algorithm (54.5% of cases, Table 4). Although this difference was not statistically significant (p = 0.22), probably due to the small number of patients with probes located outside the recommended area, this result suggests that our implantation guidelines may help to optimize probe placement.

Table 4.

Percentage of successful monitoring (microdialysis catheter located in an infarcted territory) in relation to whether placement fit new algorithm.

Probe placement	Aneurysm location	No. successful implantations (/total)	% success
Consistent with new algorithm	ACoA	15/18	83.3
	PCoA	3/4	75
	MCA	5/8	62.5
	ICA	4/4	100
	Total	27/34	79.4
Not consistent	Total	6/11	54.5

ACoA: anterior communicating artery; ICA: internal carotid artery; MCA: middle cerebral artery; N: number; PCoA: posterior communicating artery.

Justification of the need to define implantation guidelines by analysing L/P ratio changes when the probe was placed in brain area later affected by secondary infarction or not

To assess the impact of placement of the probe within secondary infarcts, we decided to analyse variations of L/P ratio by two ways. First, we analysed the median of L/P ratio over the last 12 h of monitoring in three groups (Figure 3(a)): probe located within an infarct zone (N = 33), probe outside the infarct (N = 12) and no infarct (N = 16). L/P reached 47.5 (35.0–117.3) within infarcted areas and remained close to physiological values outside the infarct (28.2 (23.2–33.0)) or in the absence of an infarct (32.6 (24.2–37.0), p = 0.002). Secondarily, we compared the numbers of hours (consecutive or not) of L/P ratio > 40 before the first scan showing infarction between the three groups (Figure 3(b)). The total duration of monitoring was similar for the three groups (219 h (148–339) no infarct, 207 h (158–295) outside infarct, 194 h (164–286) within infarct, p = 0.88). Probes located within an infarct detected a higher number of L/P > 40 events (29 (15–50)) than probes outside the infarct (4 (3–8)) or in the absence of an infarct (1 (1–5), p = 0.00001). Therefore, detection of DCI-associated neurochemical changes was far more sensitive when the probe was finally located within, as opposed to outside, the infarct zone. This suggests that it is primordial to correctly implant the probe in the brain area with the higher risk of secondary infarction.

Figure 3.

Cerebral microdialysis (CMD) data. (a) Box plots of the lactate/pyruvate (L/P) ratio median during the last 12 h of monitoring in three patient groups: (1) No detectable secondary infarction (no infarct); (2) secondary infarct apparent in scans with CMD probe outside infarct area (infarct + probe outside); and (3) secondary infarct with CMD probe within infarct area (infarct + probe within). (b) Box plots of the numbers of hours with L/P ratios > 40 before the first scan showing infarction for the groups with secondary infarction and over the whole neuro-monitoring period for the patient groups without secondary infarction. (c) Receiver operating characteristic (ROC) curve analyses of number of hours with L/P ratio > 40 for predicting long-term outcome (Glasgow outcome scale (GOS) score 1–3 vs. 4–5). Three consecutive ROC curve analyses were conducted: (1) all patients (dashed line); (2) only patients in which the probes were implanted following the new algorithm (full line) and (3) only patients with microdialysis probe within infarct area (dotted line) for examination of patient prognosis efficiency in an ideal theoretical situation; ‘ideal no misplacement’). (d) ROC curve analyses of number of hours with L/P ratio > 46 and brain tissue oxygen tension < 25 mm Hg for predicting patient outcome (same methods and representation as in panel (c)).

Influence of probe placement guidelines for predicting long-term patient outcome

To validate the interest of these implantation guidelines in practice, we analysed the potential of L/P > 40 events for predicting long-term outcome with ROC curve analyses (Figure 3(c)). When we considered all patients (N = 45 with unfavourable outcome GOS 1–3 vs. N = 16 with favourable outcome GOS 4–5), area under the curve (AUC; 0.72, < 0.8) was not sufficient to allow reliable individual prognosis. However, when we selected patients with probe placement in accordance to the algorithm (N = 32 GOS 1–3 vs. N = 14 GOS 4–5), the AUC increased (0.8), allowing reliable individual prognosis with a threshold of five L/P > 40 events (sensitivity 82.9%, specificity 71.4%). Finally, considering only probes correctly placed in a secondary infarct area (ideal case), and patients without secondary infract (N = 26 GOS 1–3 vs. N = 14 GOS 4–5), the sensitivity and specificity related to prognosis could reach 87.1% and 75%, respectively (AUC = 0.81). This would correspond to the ideal theoretical limit that prognosis could reach if probes were always implanted in the secondary infarct when it occurred.

We explored a number of possible biomarker combinations, and found that the association of L/P > 46 and PbtO₂ < 25 mmHg provided reliable biomarker of long-term outcome (Figure 3(d)). The AUC was > 0.8 when probe placement was consistent with the algorithm, and in these cases, the presence of at least two events predicted poor outcome with 85.3% sensitivity and 90% specificity (AUC = 0.87). Sensitivity and specificity were similar when we only selected patients having a probe correctly placed in a secondary infarct area or those without infarct (ideal case) (86.7% and 88.9%, respectively, AUC = 0.87). Therefore, individual patient outcome could be predicted when probe placement was optimized according to the algorithm guidelines.

We also analysed PbtO₂ data alone by univariate logistic analysis because a recent study suggested that this parameter could predict patient outcome reliably.³¹ However, in our hands, PbtO₂ was not an independent predictor of the long-term outcome (all patients: p = 0.52; patients with probe placement in accordance to the algorithm: p = 0.33; patients with probe correctly placed in a secondary infarct area or those without infarct (ideal case): p = 0.23). This difference was possibly due to the fact that we analysed the number of hours with PbtO₂ < 20 mm Hg, whereas Winkler et al. analysed the median of the daily mean baseline values of PbtO₂ over the whole monitoring period. In addition, our clinical practice led us to treat low PbtO₂ (<15 mm Hg) by increasing of the fraction of inspired oxygen, by way of normobaric oxygen therapy or by cerebral perfusion pressure optimization, which could have confounded our approach.

Discussion

In this study, ruptured aneurysms produced a consistent pattern of secondary infarct zones that most often impacted an ACA territory for the ACoA aneurysms and ipsilateral MCA territory for the PCoA, ICA and MCA aneurysms. Hence, to maximize the likelihood of monitoring a secondary infarct, local probes should be implanted in these respective brain areas.

We estimated, retrospectively, the probability of monitoring from a secondary infarct (i.e. successful monitoring) in patients who had received CMD/PbtO₂ monitoring. Following the specific guidelines, as described above, increased the probability of probe placement within a brain area that would later be impacted by an infarct from 54% to 79%.

By analysing CMD/PbtO₂ data obtained from within versus outside of infarct areas, we obtained support for the assumption that detection of DCI-associated neurochemical changes is more sensitive when the probe is finally located within the infarct zone. Our results suggest that these neurochemical changes are limited to the area of infarct, and not detectable by probes located away from the injury site, underscoring the importance of adopting guidelines that maximize the likelihood of placing probes within a secondary infarct zone.

Few studies have investigated the correlation between aneurysm location and DCI occurrence. In a study of 56 patients, Rabinstein et al.¹¹ observed two patterns of focal lesions near the aneurysm rupture and multiple widespread lesions. In a study of 29 patients, Miller et al.³² described frequent secondary ischemic lesions in the ACA territory following ACoA aneurysms, and in the ipsilateral MCA territory following non-midline aneurysms. Finally, Ulrich et al.²⁴ determined the probability that a secondary infarct would affect the area monitored in a cohort of 100 patients with ASAH equipped with a focal oxygen probe to validate retrospectively the Bern/Frankfurt algorithm (see Table 2). Our results are largely confirmatory of these data in a larger population, including PCoA aneurysm cases that could not be analysed previously due to the small numbers of patients, and extends these previous analyses.

The current algorithm and the Bern/Frankfurt algorithm²⁴ are refinements of the imprecise 2004 consensus guidelines,²¹ providing more precise territories for probe implantation (see Table 2 for guideline comparison). An important distinction between the two lies in the guidelines for aneurysms of the posterior circulation. In these cases, the parent vessel territory is difficult to access in comatose patients lying in dorsal decubitus. The Bern/Frankfurt algorithm²⁴ recommends implantation in the MCA territory for all aneurysms of the posterior circulation. In our hands, this approach yielded a 47% success rate, and in Ulrich et al.’s study,²⁴ only a 23% success rate. Thus, the risks and benefits of implanting a probe with a less than 50% chance of successfully monitoring from an infarcted area should be weighed very carefully.

Invasive cerebral monitoring techniques are accepted by the neurocritical care consensus.^22,25 Here, we observed that a new implantation algorithm improved individual outcome prediction. CMD/PbtO₂ monitoring may support long-term outcome predictions by revealing neurobiological patterns observed before imaging reveals a secondary infarction. Such individual prognosis could be instrumental in directing specific therapeutic interventions, such as haemodynamic/volemic optimization or angioplasty,^25,33 in accordance with the individual needs of each patient.

This study has the limitation of being based on retrospective analyses. Thus, the reliability of these guidelines should be confirmed in a prospective study. Moreover, the timing of probe implantation directly affects the number of hours during which the L/P threshold was exceeded. However, in our study, this delay was controlled as the probe was implanted in the third day. It would thus appear necessary, in the prospective confirmatory study, that the timing of probe implantation and the ensuing monitoring period should be respected to correctly determine the number of hours of excessive L/P ratios necessary for the prediction of patient outcome or for DCI detection. We also acknowledge the invasive characteristics of this type of cerebral monitoring and have not had the possibility of comparing it with data from more conventional, less invasive monitoring, such as transcranial Doppler and Near-Infrared Spectroscopy (NIRS) brain saturations. The association of transcranial Doppler velocities and NIRS proximal probe saturations with CMD/PbtO₂ values could possibly allow for a secondary validation of correct probe placement. In addition, although our approach does not demonstrate a benefit for the patient of detecting DCI early using invasive probes, it paves the way for future interventional studies, in which therapeutic intervention could be directed by information gathered via intracerebral probes.

In summary, we suggest an algorithmic approach to probe placement leading to improved detection potential for DCI and to increase the performance of outcome prediction with neurochemical changes. Cerebral monitoring probes should be implanted within an ACA territory for ACoA aneurysms and within the ipsilateral MCA territory for MCA, ICA and PCoA aneurysms. These guidelines are easily implemented in clinical practice and a more precise monitoring may lead to more informed therapeutic decisions and better patient outcome.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Inserm U1028, CNRS UMR5292, University Claude Bernard Lyon I and by Hospices Civils de Lyon.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

YT, GKB, FD, APL and SM contributed significantly to the overall work design. YT, GKB, FD, EO, EC, SG and TL contributed to the acquisition of data and YT, SM and EO interpreted the data for the draft of the article. YT and SM drafted the article and GKB, APL, RC, TL and FD revised it critically. All the authors approved this version to be published.