Dose Optimization Study of Therapeutic Hypercapnia for Prevention of Secondary Ischemia After Severe Subarachnoid Hemorrhage


 BackgroundAim of this study was to investigate the time point at which the cerebral blood flow (CBF) enhancing effect of controlled hypercapnia in patients with severe aneurysmal subarachnoid hemorrhage (SAH) starts to extenuate. This point is assumed to be the time at which buffer systems become active and annihilate a possible therapeutic effect. MethodsIn this prospective interventional study in a neurosurgical ICU the arterial partial pressure of carbon dioxide (PaCO2) was increased to a target range of 50 - 55 mmHg for 120 minutes by modification of the respiratory minute volume (RMV) one time a day between day 4 and 14 in 12 mechanically ventilated poor-grade SAH-patients. Arterial blood gases were measured every 15 minutes. CBF and brain tissue oxygen saturation (StiO2) were the primary and secondary end points. Intracranial pressure (ICP) was controlled by an external ventricular drainage. ResultsUnder continuous hypercapnia (PaCO2 of 53.17 ± 5.07), CBF was significantly elevated between 15 and 120 minutes after the start of hypercapnia. During the course of the trial intervention, cardiac output also increased significantly. To assess the direct effect of hypercapnia on brain perfusion, the increase of CBF was corrected by the parallel increase of cardiac output. The maximum direct CBF enhancing effect of hypercapnia of 31% was noted at 45 minutes after the start of hypercapnia. Thereafter, the CBF enhancing slowly declined. No relevant adverse effects were observed. Conclusion CBF and StiO2 reproducibly increased by controlled hypercapnia in all patients. After 45 minutes, the curve of CBF enhancement showed an inflection point when corrected by cardiac output. Temporary hypercapnia of 45 minutes is, thus, likely to be the optimum duration for a therapeutic use and for a controlled comparative trial. Longer intervals bear the risk of a negative rebound effect after return to normal ventilation parameters and may be counterproductive inducing ischemia in a state of critical perfusion after SAH. Trial registrationThe study was approved by the institutional ethics committee (AZ 230/14) and registered at ClinicalTrials.gov (Trial-ID: NCT01799525). Registered 01 January 2015. Retrospectively registered.


Abstract Background
Aim of this study was to investigate the time point at which the cerebral blood ow (CBF) enhancing effect of controlled hypercapnia in patients with severe aneurysmal subarachnoid hemorrhage (SAH) starts to extenuate. This point is assumed to be the time at which buffer systems become active and annihilate a possible therapeutic effect.

Methods
In this prospective interventional study in a neurosurgical ICU the arterial partial pressure of carbon dioxide (PaCO 2 ) was increased to a target range of 50 -55 mmHg for 120 minutes by modi cation of the respiratory minute volume (RMV) one time a day between day 4 and 14 in 12 mechanically ventilated poor-grade SAH-patients. Arterial blood gases were measured every 15 minutes. CBF and brain tissue oxygen saturation (StiO 2 ) were the primary and secondary end points. Intracranial pressure (ICP) was controlled by an external ventricular drainage.

Results
Under continuous hypercapnia (PaCO 2 of 53.17 ± 5.07), CBF was signi cantly elevated between 15 and 120 minutes after the start of hypercapnia. During the course of the trial intervention, cardiac output also increased signi cantly. To assess the direct effect of hypercapnia on brain perfusion, the increase of CBF was corrected by the parallel increase of cardiac output. The maximum direct CBF enhancing effect of hypercapnia of 31% was noted at 45 minutes after the start of hypercapnia. Thereafter, the CBF enhancing slowly declined. No relevant adverse effects were observed.
Conclusion CBF and StiO 2 reproducibly increased by controlled hypercapnia in all patients. After 45 minutes, the curve of CBF enhancement showed an in ection point when corrected by cardiac output. Temporary hypercapnia of 45 minutes is, thus, likely to be the optimum duration for a therapeutic use and for a controlled comparative trial. Longer intervals bear the risk of a negative rebound effect after return to normal ventilation parameters and may be counterproductive inducing ischemia in a state of critical perfusion after SAH.

Introduction
Cerebral vasospasm and delayed cerebral ischemia (DCI) are much-feared complications and the main reason for prolonged hospitalization and persistent disabilities after aneurysmal subarachnoid hemorrhage (SAH). Cerebral blood ow (CBF) is -under physiological conditions -regulated by autoregulation as reactivity to changes of transmural wall tension, by equilibrium of vessel relaxing and constricting factors and by changes of the arterial partial pressure of carbon dioxide (PaCO 2 ). While autoregulation is deranged and the balance of paracrine factors shift towards vessel contraction, the PaCO 2 reactivity may be more or less altered but is, in principle, intact after aneurysmal SAH (1,2). The degree of reactivity of CBF after SAH has been shown to be an indicator for the course of the disease after SAH with poor reactivity indicating a high risk for delayed cerebral ischemia and poor prognosis.
During delayed vasospasm of proximal intracranial vessel trunks, hyperventilation induces ischemia and infarction due to an additional constriction of small distal blood vessels and is associated with poor outcome after SAH (3,4). However, previous trials investigated only minor PaCO 2 variations around the normal range. Petridis and co-workers have demonstrated that permissive hypercapnia is not hazardous to patients after SAH if they have an external ventricular drainage (EVD) continuously draining cerebrospinal uid (CSF) during the period of hypercapnia (5). In a previous trial, it was demonstrated that more pronounced changes of PaCO 2 can reproducibly affect CBF even after severe SAH suggesting a therapeutic potential of temporary hypercapnia (6). However, it is well known that alterations of arterial blood gases are followed by adaptation mechanisms. Research data on high altitude adaptation showed that the effect of increased CBF due to hypercapnia is timely limited (7,8).
The optimum duration of controlled hypercapnia and its effects on CBF and brain tissue oxygen saturation (StiO 2 ) as a therapeutic tool against DCI are not clari ed. In this prospective trial we aimed to investigate the course of CBF during longer-term hypercapnia and determine the time-point at which physiologic adaptation mechanisms start to mitigate the CBF-increasing effect in order to locate the optimum duration of controlled hypercapnia in poor-grade SAH patients.

Material And Methods
The study was approved by the institutional ethics committee (AZ 230/14) and registered at ClinicalTrials.gov (Trial-ID: NCT01799525). The study has been conducted in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Written consent of a legal guardian was obtained for each patient prior to inclusion into the study.

Inclusion criteria
Patients were eligible for recruitment if they had suffered severe aneurysmal SAH (Hunt/Hess grade 3-5, Fisher grade 3) no longer than 96 hours ago, had received external ventricular drainage (EVD) due to occlusive hydrocephalus and early treatment of the ruptured aneurysm (coiling or clipping). If no other condition like pulmonary complications after aspiration or elevated intracranial pressure prohibited a reduction of analgosedation after occlusion of the aneurysm, an attempt was made to reduce analgosedation. If the patient did not show appropriate reactions, anaesthetics were reapplied and the patient was included in the study.

Exclusion criteria
Age under 18 years, pregnancy, non-aneurysmal SAH and the presence of a not occluded aneurysm were exclusion criteria. Furthermore, patients suffering from chronic obstructive lung disease (COLD) were excluded because their cerebrovascular reactivity to changes of PaCO 2 in the target range of this study is unclear. Patients eligible for inclusion who had an ICP over 20 mmHg were not included and were reevaluated later for a possible start of the study intervention.

Safety measures and criteria for an interruption of the study intervention
Study interventions were performed in 24-hour intervals (24 ± 1 hour). Prior to the daily study intervention, cardiovascular parameters, blood gases and ICP were assessed. If resting ICP was over 20 mmHg at the desired timepoint, the study intervention was not conducted and the patient re-assessed 2 hours later. If ICP was still elevated over 20 mmHg, the study intervention was not performed on that day. Criteria to interrupt the trial intervention were prede ned as an increase of ICP over 25 mmHg for more than two minutes, hypoxemia (PaO2 < 70 mmHg or sO2 < 90 mmHg) and severe arterial acidosis with an arterial pH-value under 7.25.

Termination of the study procedures
Study interventions were terminated on day 14. Prede ned criteria to terminate the study procedures in an individual patient before day 14 were the cessation of therapy due to a poor prognosis of whatever reason according to the treating physicians' estimation and if a reduction of analgosedation was possible and the study intervention was no longer feasible due to spontaneous breathing.

Respiratory settings and study intervention
All patients were intubated and mechanically ventilated as de ned by the inclusion criteria. If not indicated for ventilatory reasons anyway, the respirator was brought into a volume-controlled ventilation mode for the daily study intervention (Servo-i, Maquet, Rastatt, Germany) maintaining the same respiratory minute volume (RMV). After a rst arterial blood gas analysis (RapidLab, Siemens Health Care, Erlangen, Germany), respirator settings were adjusted to reach a PaCO 2 within the normal range (36-44 mmHg). This was con rmed by another blood gas analysis. The respiratory rate was then reduced to lower the RMV by 40%. This value was adopted from a previous study assessing the feasibility and risk pro le of the study intervention (6) and served as a guide value for further respirator adjustments to reach the target PaCO 2 of 50-55 mmHg. Arterial blood gases were then measured in 5minute intervals and ne adjustment of the respirator settings made until a steady PaCO 2 between 50-55 mmHg was reached. Thereafter, arterial blood gases were measured in 15-minute intervals and further respirator adjustments made in order to maintain a stable PaCO 2 of 50-55 mmHg for 120 minutes. All other respiratory settings (tidal volume, inspiratory-expiratory ratio, and inspiratory fraction of oxygen) remained unchanged. Respirator settings were returned to pre-trial settings to obtain normal PaCO 2 levels after 2 hours.

Target Parameters
Parameters measuring CBF and brain tissue oxygenation were the prede ned study endpoints.
Direct measurement of CBF: The course of CBF, was measured by an intracerebral thermodilution probe (Q-Flow 500®) located in the right frontal cortex. The probe was placed 1.5 cm anterior to the external ventricular drainage and was monitored by a Bowman Perfusion Monitor® (Hemedex™, Cambridge, MA, USA). To obtain a continuous measurement and avoid recalibration during the study intervention, the automatic calibration was paused during the time of the study intervention (9). Transcranial Doppler sonography (TCD): Prior to the beginning of each study intervention, baseline measurements of TCD were made. At each time-point during the study intervention, TCD measurements were repeated.
Measurements were conducted by a well-trained medical technician otherwise not involved in the study procedures with an experience of over 15 years of TCD measurements.
Cerebral tissue oxygen saturation (S ti O 2 ): S ti O 2 was measured by bilateral near-infrared spectroscopy (NIRS) electrodes (INVOS®, Covidien, Neustadt/Donau, Germany), which were attached on both sides of the forehead. A baseline value was set prior to the intervention to which the following measurements (every 15 minutes for 2 hours) were compared.

Physiological Parameters
All patients were treated on a neurointensive care unit and received the routine continuous monitoring of cardiovascular parameters including electrocardiography, invasive blood pressure monitoring, monitoring of intracranial pressure via an EVD and central venous pressure (CVP). In patients included into the study, a pulse contour-based measurement of cardiac output (CO) was added to the invasive blood pressure monitoring (proAQT, Pulsion Medical Systems SE, Munich, Germany) in order to follow changes of CO during hypercapnia.
Protocol and Statistical Analysis CBF, S ti O 2 , ICP, arterial blood gases, pH and hemodynamic parameters were measured at baseline and at 0, 15, 30, 45, 60, 75, 90, 105 and 120 minutes of the intervention and documented in a paper-based examination protocol. Acquired data was transferred to Excel®-Worksheets (Microsoft®, Redmont, WA, USA). Dynamic changes over time were analysed using a one-way ANOVA for repeated measures. In case of statistical signi cance, time-points were compared to baseline values using a Dunnett's post hoc test.
Statistical signi cance was de ned if p < 0.05. Statistical analysis was performed using GraphPad Prism 4.0 Statistical Software (GraphPad, La Jolla, CA, USA).

Results
All 12 patients included in this trial suffered from severe aneurysmal SAH. The average and median Hunt/Hess grade was 4 with four patients classi ed as Hunt/Hess grade 3, four patients as Hunt/Hess grade 4, and four patients as Hunt/Hess grade 5. All patients had thick subarachnoid blood layers and clots according to grade 3 of the Fisher scale, in ten patients additional intracerebral and/or intraventricular blood was detected in the initial CT scan. Cerebral angiogram at admission revealed a total of 15 aneurysms (3 patients with two aneurysms), of which 13 were occluded within 96 hours after ictus, 6 by surgical clipping, and 7 by endovascular coiling. The remaining two aneurysms were not considered to be the origin of SAH in patients with more than one aneurysm. Mean age at admission was 48 ± 7.93 years, 9 were female and 3 were male. Patient characteristics, hemorrhage-related features and location and treatment of the ruptured aneurysms are depicted in Table 1. A total of 106 trial interventions were performed from day 4 to day 14 after SAH. Table 1 Personal, hemorrhage-related and aneurysm-related characteristics of 12 study patients. 2 patients developed minor secondary infarction (one partial left MCA and one partial right ACA territory). One patient with primary large right temporal ICH developed territorial infarctions in both MCA territories in the course of treatment. Two patients died, one from rebleeding from an endovascularly occluded aneurysm outside of study procedures and one due to extended infarction in both MCA territories. (ICH = intracerebral hemorrhage, IVH = intraventricular hemorrhage, ICA = internal carotid artery, MCA = middle cerebral artery, Acom = anterior communicating artery, GOS = Glasgow Outcome Scale) Clinical and radiological course Mean initial Glasgow Coma Score (GCS) was 8 with deterioration to a GCS of 6 prior to intubation. All patients had developed occlusive hydrocephalus and received an external ventricular drainage (EVD).
After occlusion of the aneurysm anesthetics were withdrawn and the patients were allowed to wake up. If regain of consciousness was not appropriate and extubation not possible, anesthetics were readministered and the patients were included in the study after written informed consent was given by a legal guardian. 11 of 12 patients showed a signi cant increase in daily transcranial Doppler sonography (TCD) and/or Perfusion-CT and underwent follow-up angiography.

Ventilation and blood gas analysis
To reach a target PaCO 2 of 55 mmHg, the baseline RMV was reduced by 40%. After the rst arterial blood gas control, the respirator settings were further adjusted (Fig. 1). Prolonged hypercapnia led to a decrease in pH as a sign of respiratory acidosis. Nevertheless, no patient reached a pH under 7.25 as a criterion for trial interruption. Mean pH was 7.44 with normocapnia before trial interventions and decreased to 7.28 after 75 minutes with a slight increase to 7.29 after 90 minutes, reaching a steady state of 7.29 at 90 to 120 minutes of hypercapnia (Table 2). Table 2 Arterial blood gas (ABG) values in 15-minute intervals after the start of the trial intervention. After measurement of the baseline value, the respiratory minute volume (RMV) was reduced. The rst ABG analysis after reducing the RMV was taken 5 minutes thereafter to exclude an exaggerated individual effect of the reduction of RMV. "Hyp 0 min" indicates a timepoint 15 minutes after the start of the study intervention and reduction of RMV. This value was then used to further modify the RMV to reach a target of 55 mmHg.
(PaCO 2 = arterial partial pressure of carbon dioxide, PaO 2 = arterial partial pressure of oxygen, Hyp + # min indicates the time after the start of the hypercapnic trial intervention) Intracranial pressure, arterial blood pressure, cerebral perfusion pressure Intracranial pressure (ICP) increased slightly but signi cantly within the rst 45 minutes. Baseline ICP before the start of the study intervention was 11.7 ± 2.4 mmHg at a PaCO 2 of 38.8 ± 3.7 mmHg. The maximum ICP value of 12.7 ± 3.0 mmHg was recorded immediately after induction of hypercapnia at a PaCO2 of 47.1 ± 4.8. Thereafter, a surplus CSF collection resulted in a continuous decline of ICP thereafter to a nal value of 11.8 ± 2.6 mmHg after 120 minutes (Fig. 2). Mean arterial blood pressure (MABP) increased from a baseline of 100.4 ± 9.6 mmHg to a maximum of 102.2 ± 11.3 mmHg immediately after the induction of hypercapnia and continuously decreased thereafter to reach a nal value of 96.0 ± 8.0 mmHg after 120 minutes. Baseline CPP was 88.6 ± 9.9 mmHg to increase to a maximum of 89.5 ± 11.2 mmHg after induction of hypercapnia and continuously decrease to 84.2 ± 7.9 mmHg at the end of the monitoring time. The decreases of MABP and CPP both were signi cant from 60 minutes until 120 minutes after induction of hypercapnia.
Cerebral Tissue Oxygenation S ti O 2 measured over the right forebrain increased to a maximum of 109 ± 8.8% of baseline 60 minutes after induction of hypercapnia at a PaCO 2 of 54.22 ± 5.27 mmHg. Over the left forehead, S ti O 2 increased to a maximum 109 ± 7.9% of baseline after 60 minutes into the trial intervention, respectively. The course of S ti O 2 during hypercapnia is depicted in Fig. 3.

Transcranial Doppler sonography
Mean ow velocities (MFV) in TCD increased in both middle cerebral arteries during the trial intervention. In the right middle cerebral artery (MCA), MFV increased from a baseline of 112 ± 35 cm/s to a steady level of 140-145 cm/s between 45 and 120 minutes into the trial intervention. In the left MCA, MFV increased from a baseline of 106 ± 51 cm/s to values between 135 and 140 cm/s from 45 to 120 minutes after induction of hypercapnia. The increase was signi cant throughout the entire period of monitoring ( Fig. 4).

Cardiac Output
During the trial intervention, cardiac output (CO) increased from a baseline of 6.3 ± 2.0 l to 6.6 ± 2.0 l (106%) after induction of hypercapnia, 7.1 ± 2.3 l (114%) after 45 minutes and 7.5 ± 2.5 l (120%) after 120 minutes. This increase was caused by an equal 10% increase of stroke volume and heart rate.

Cerebral Blood Flow
Increasing PaCO 2 from a baseline value of 38.77 ± 3.75 mmHg to 53.18 ± 7.16 mmHg after 120 minutes resulted in a nal increase of cerebral blood ow (CBF) to149 ± 93% of baseline. After 30, 45, 60, 75 and 90 minutes CBF increased to 141 ± 92%, 146 ± 99%, 140 ± 80%, 144 ± 80%, 151 ± 86%, and 147 ± 93% of baseline values, respectively. The peak value was recorded after 75 minutes (Fig. 5). In order to calculate the intervention's direct effect upon the brain vasculature, the increase of CBF was corrected by the concomitant increase of CO resulting in a maximum net increase of CBF to 132% of baseline after 45 minutes and decreased thereafter to reach a nal value of 129% of baseline at the end of the monitoring period (Fig. 6).

Discussion
In a previous phase 1 trial, it was observed that CBF can be reproducibly increased by intermittent controlled hypercapnia in the days following aneurysm rupture in patients with poor grade SAH (10,6). In that trial, CBF increased during a sequential increase of arterial PaCO 2 . After resetting mechanical ventilation to baseline parameters, CBF showed a slow and asymptotic return to starting levels without a negative rebound effect. This observation suggested that a longer duration of hypercapnia might even extend the CBF-increasing effect. The present study was planned as a dose optimization study in order to identify the time-point at which CBF reaches a maximum. It was assumed that after a maximum CBF increase upon elevated PaCO 2 , buffer mechanisms in blood and CSF may result in adaptation mechanisms that lead rst to an attenuation of the CBF-increasing effect and then to a negative rebound effect after the hypercapnic challenge is terminated. For safety reasons, a preliminary termination of the study intervention was included into the study protocol if the CBF-enhancing effect secondarily declined by more than 25%. In order to identify the time-point of the maximum intrinsic effect of controlled hypercapnia upon the cerebral vasculature, the value was corrected by the possible inotropic effect of prolonged hypercapnia which has been described previously (11). Indeed, a positive inotropic effect was also noted during the study interventions conducted in this present study, moderate but statistically signi cant. It resulted in a "net" optimum effect of the hypercapnic challenge on the cerebral vasculature of 45 minutes. After the proof of principle in a previous study and the present results we suggest that this may be the basis for an evaluation of e cacy in a controlled clinical trial.
Aneurysmal SAH is still a life-threatening incident with poor over-all prognosis. Its course is characterized by serially occurring ischemic events. Immediately after the rupture of the aneurysm, cerebral perfusion pressure (CPP) decreases due to a sudden increase of ICP resulting in a reduction of CBF and global ischemia. Simultaneously or shortly thereafter, diffuse early arterial vasoconstriction has been observed and contributes to the persistent perfusion de cit in the early stage of SAH (12,13). Due to developments in emergency care, an increasing number of patients with poor-grade SAH can be adequately resuscitated and nd their way to a specialized unit. Patients in a poor clinical condition and large amounts of blood in the subarachnoid space, in turn, are prone to develop vasospasm and delayed cerebral ischemia. To date, no single drug or manipulation has proven undoubtedly effective to prevent or treat the risk for DCI. Even the evidence for nimodipine, the most widely used pharmacological attempt to prevent DCI is based on a single trial which was conducted more than 30 years ago before endovascular aneurysm and vasospasm therapy was invented and without intensive care therapy with contemporary standards (14,15).
Delayed vasospasm after a few days, caused by endothelial dysfunction, structural changes of the vessel wall, and in ammatory processes is likely to be an important factor for delayed ischemic neurological de cits (DIND) and delayed cerebral ischemia (DCI) (16,17). A variety of clinical and experimental studies have investigated various approaches to prevent vasospasm (14,18,19,20). But even initially promising trials, e.g. with endothelin A receptor antagonists, successful in treating vasospasm, did not prevent DCI or improve clinical outcome (21).
From a pathophysiological point of view, delayed ischemia after SAH is a slowly arising perfusion de cit rather than a sudden and complete ischemia. Energy depletion, therefore, is also likely to be gradual. Thus, the rationale behind intermittent controlled hypercapnia is to temporarily increase a critically reduced CBF so that energy stores can recover.
It has long been shown that cerebral autoregulation of arterial blood pressure is disturbed after severe aneurysmal SAH and that the equilibrium of local factors are shifted towards constriction. However, the reactivity to changes of PaCO 2 remains vital (2,3,22,23). Increased PaCO 2 may not have an effect on CBF and vessel diameter in the early stage of SAH as recently published by Friedrich and co-workers (24). For the chronic phase, however, previous work has shown a highly reproducible increase of S ti O 2 and CBF under staged hypercapnia. Since this effect was still reproducible during angiographically proven vasospasm (6) the concept of maximum peripheral vessel dilation downstream of large-trunk vasospasm may have to be rethought. At the same time, it gives space for a therapeutic intervention. A previous phase 1 study has not only shown the proof of principle without noteworthy negative side effects, but also a surprising reduction of the incidence of DCI and relatively favorable outcome as compared to historical controls and data published in literature (6,10).
The previous results make it seem worthwhile to evaluate this procedure regarding its therapeutic effect. However, to date no data about the ideal duration of hypercapnia on CBF and cerebral tissue oxygenation exists. From a physiological point of view these positive effects must be temporary since buffer mechanisms in blood and CSF, connected via the carboanhydrase enzyme, trigger an adaptation to longterm changes of CO 2 .To the best of our knowledge, there is no literature about the long-term effects of hypercarbia upon CBF and the course of its adaptation. From high altitude research it is known that mountaineers start to hyperventilate during ascent, which results in a decrease of CBF. After several hours or days, depending on the altitude, acclimatization processes are leading to a normalization of ventilation and physiological parameters (7,8). However, this is not the timeframe of interest for the particular needs of our patient collective. In contrast to our study protocol the effects in high altitude and its impact on CBF do not occur in a range of seconds or minutes but rather within hours and days during or after and ascent. An additional factor is the lower partial pressure of O 2 in high altitude leading to hypoxia; therefore, this data is not comparable with normoxic conditions on an intensive care unit. Adaptation mechanisms to long-term hypercapnia, e.g. buffer in the cerebrospinal uid or blood via carboanhydrase are neither well known nor investigated in trials up to now. Alterations of the pH of the brains' extracellular space as a regulator of the cerebrovascular response to CO 2 were discussed by Lassen et al., but remained unproven (25). Raichle and co-workers examined the effects of transient hyperventilation for several hours on CBF in healthy volunteers covering the period that is of interest for our particular study setting. They showed that CBF decreased with the onset of hyperventilation, but gradually recovered again under continuous hyperventilation and nally exceeded baseline levels in terms of a rebound effect when ventilation was returned to normal. PaCO 2 also decreased initially but remained constantly low under hyperventilation and returned to baseline parameters after ending of hyperventilation (26).
We assume that the time-course, extent and reaction upon adaptation to hypercapnia is approximately the same as the adaptation to hyperventilation as reported by Raichle et al. (26). This implies that in patients with SAH, whose CBF is critically reduced, there is the danger that a negative rebound effect may carry CBF below ischemic thresholds even if there is a bene cial effect during the study intervention.
Therefore, the duration of CBF hypercapnia must not exceed a safe time-frame.
Previous data on the effect of graded hypercapnia after aneurysmal SAH (6)  As a result of disturbed cerebral autoregulation after SAH, systematic increase of cardiac output alone, e.g. via application of catecholamines, could lead to an increase in CBF. Adjusting CBF by cardiac output, the hypercapnia-induced net increase of CBF was highest as early as 45 minutes after the start of the trial intervention. Given the statistically signi cant elevation of cardiac output after 60 minutes with further increase under prolonged hypercapnia, sustained therapeutic hypercapnia after reaching its maximum effect on CBF seems not advisable; adaptation mechanisms and negative cardiac effects bear the danger of a negative rebound effect with a secondary decline of CBF below the starting level.

Conclusion
As a result of this dose optimization study we conclude that 45 minutes of controlled hypercapnia is the most suitable duration regarding the CBF-increasing effect and the patients' safety. Considering that a complete normalization of cardiac and physiologic parameters is achieved several hours after returning to baseline ventilation parameters it might be even more bene cial to repeat the CO 2 -intervention at higher frequency, e.g. 2 or 3 times a day, although this has not been explicitly tested in the current trial. A randomized, controlled trial with 45 minutes of controlled hypercapnia at 12-hour intervals is planned for further evaluation of the therapeutic e cacy of this method in poor-grade aneurysmal SAH.

Declarations
Ethics approval and consent to participate The study was approved by the institutional ethics committee (AZ 230/14) and registered at ClinicalTrials.gov (Trial-ID: NCT01799525). Written consent of a legal guardian was obtained for each patient prior to inclusion into the study.

Consent for publication
Written consent of a legal guardian was obtained.

Figure 2
Course of intracranial pressure (ICP) during the study intervention. All patients had an external ventricular drainage which was allowed to continuously drain cerebrospinal uid during the intervention. No relevant increase of ICP was observed during the monitoring period. The course of cerebral blood ow (CBF) during the study intervention without measured by a right frontal intracerebral thermodilution probe. (*p < 0.05, **p < 0.01, ***p < 0.001)