Association of abnormal carbon dioxide levels with poor neurological outcomes in aneurysmal subarachnoid hemorrhage: a retrospective observational study

Background In patients with aneurysmal subarachnoid hemorrhage (SAH), an association between hypocapnia and poor clinical outcomes has been reported. However, the optimal arterial carbon dioxide tension (PaCO2) remains unknown. The present retrospective study aimed to examine the association of abnormal PaCO2 levels with neurological outcomes and investigate the optimal target PaCO2 level in patients with SAH. Methods We retrospectively selected consecutive adult patients hospitalized in the intensive care unit (ICU) for SAH between January 2009 and April 2017. Univariate and multivariate analyses were performed to identify the independent predictors of unfavorable neurological outcomes (i.e., modified Rankin scale score of 3–6 on hospital discharge). Results Among 158 patients with SAH, 73 had unfavorable neurological outcomes. During the first 2 weeks in the ICU, the median number of PaCO2 measurements per patient was 43. The factors significantly associated with unfavorable neurological outcomes were age, Hunt and Kosnik grade, maximum lactate levels during the first 24 h, and maximum (odds ratio [OR], 1.12; 95% confidence interval [CI], 1.03–1.21; p < 0.01) and minimum PaCO2 levels (OR, 0.81; 95% CI, 0.72–0.92; p < 0.01). Receiver operating characteristic curve analysis revealed that the cutoff range of PaCO2 was 30.2–48.3 mmHg. Unfavorable neurological outcomes were noted in 78.8% of patients with PaCO2 levels outside this range and in 22.8% of patients with PaCO2 levels within this range. Conclusions Both the maximum and minimum PaCO2 levels during ICU management in patients with SAH were significantly associated with unfavorable neurological outcomes. Further prospective studies are required to validate our findings and explore their clinical implications. Our findings may provide a scientific rationale for these future prospective studies. Electronic supplementary material The online version of this article (10.1186/s40560-018-0353-1) contains supplementary material, which is available to authorized users.


Background
Aneurysmal subarachnoid hemorrhage (SAH) accounts for a reasonably large proportion of stroke-related mortality cases. The overall mortality in patients with SAH is over 30%, and approximately 10-20% of survivors show functional dependence despite intensive neurological care [1,2]. Several extensive studies have been conducted to improve intensive neurological care in patients with SAH [3][4][5][6][7][8].
Cerebral blood flow (CBF) is mainly regulated by arterial carbon dioxide tension (PaCO 2 ) [9,10]. Abnormal PaCO 2 levels are considered to cause major changes in CBF through vasoconstriction and vasodilation, respectively, possibly contributing to further brain injury [11,12]. Additionally, a previous systematic review demonstrated abnormal PaCO 2 levels to be associated with poor clinical outcomes in patients with traumatic brain injury (TBI), post-cardiac arrest patients, and stroke patients [13]; therefore, PaCO 2 control in the intensive care unit (ICU) can greatly influence the management of patients with SAH. Several studies have reported that hypocapnia (defined as PaCO 2 < 35 mmHg), albeit unintentional, may be a common and under-recognized cause of brain tissue hypoxia after SAH and may be associated with poor neurological outcomes and a high incidence of delayed cerebral ischemia (DCI) [14][15][16]. However, the association between hypercapnia (defined as PaCO 2 > 45 mmHg) and neurological outcomes in patients with SAH has not been examined.
The present guidelines for the management of SAH do not specify the target PaCO 2 level [17,18]. Recently, a phase I study involving poor-grade SAH patients demonstrated that CBF increased after controlled hypercapnia by up to 60 mmHg, even during periods of vasospasm [10]. Another study involving resuscitated cardiac arrest patients reported that the levels of neuron-specific enolase were lower with targeted therapeutic mild hypercapnia (PaCO 2 , 50-55 mmHg) than with normocapnia (PaCO 2 , 35-45 mmHg) [19]. Therefore, it is unknown whether hypercapnia is permitted and whether there is an optimal target PaCO 2 level in patients with SAH.
The present study aimed to examine the association of abnormal PaCO 2 levels with unfavorable neurological outcomes and investigate the target PaCO 2 level in patients with SAH for further prospective studies.

Study design and setting
This single-center, retrospective, observational study was performed at Kagawa University Hospital, a 613-bed teaching institution with an 8-bed ICU managed by a neurointensivist. The patient medical records were reviewed after receiving approval from the institutional review board (approval number H29-090), and the study was performed in accordance with the ethical standards established in the 1964 Declaration of Helsinki and its later amendments. The institutional review board approved a waiver of the requirement for individual consent. If patients did not want to participate in the current study, they could request to be excluded.

Study participants and inclusion criteria
The study included patients aged > 18 years who were consecutively admitted to the ICU with a confirmed diagnosis of SAH between January 1, 2009, and April 15, 2017, and who had at least five available arterial blood gas (ABG) analyses. The exclusion criteria were a history of pregnancy or trauma, presence of acute lung injury/ acute respiratory distress syndrome or severe heart failure requiring positive-pressure ventilation during the first 2 weeks in the ICU, and provision of only comfort care ( Fig. 1).

General management of SAH in the ICU
All the patients with SAH were managed in accordance with the Guidelines for the Management of Aneurysmal Subarachnoid Hemorrhage by the American Heart Association/American Stroke Association [8,18]. In addition to general intensive care, all the patients were monitored for clinical deterioration or for the development of cerebral infarction owing to DCI. Using previously published criteria, DCI was defined as either (1) the development of new focal neurological signs or deterioration of the consciousness level thought to be caused by the presence of ischemia after the exclusion of other possible causes or (2) the appearance of new infarcts caused by vasospasm on CT or MRI [15,20]. Fluid management was targeted to maintain euvolemia. Induced hypertension and hemodilution were not performed, and mannitol or hypertonic saline solution was not administered. The patients received minimum amounts of sedatives, such as propofol, midazolam, and dexmedetomidine, which were necessary to prevent ventilator dyssynchrony and patient discomfort. Analgesics, including acetaminophen, nonsteroidal anti-inflammatory medications, and fentanyl, were administered as required. Fever was treated aggressively with acetaminophen, nonsteroidal anti-inflammatory medications, or cooling devices. Active maintenance of normothermia was not routinely performed.
Control of PaCO 2 levels in patients with SAH in the ICU ABG analyses, including PaCO 2 , were routinely performed every 6 h during the first 2 weeks in the ICU, and additional measurements were obtained by critical care physicians as needed. Patients were typically managed in the ICU during the first 2 weeks with reference to ABG analyses data, considering the timing of DCI occurrence. For patients in whom the arterial line could be withdrawn and who could be discharged from the ICU at an early stage, periodic ABG monitoring was completed by day 14. All patients were intubated prior to angiography and the initial treatment (surgical clipping or endovascular coiling) using a muscle relaxant. The muscle relaxant was discontinued after the completion of the initial treatment and was not used as ICU management thereafter. In mechanically ventilated patients, respiratory management usually involved the use of the spontaneous breathing mode (pressure support mode) with minimal support (i.e., continuous positive airway pressure or a pressure support level ≤ 5 cm H 2 O). Daily spontaneous breathing trials were not routinely performed. When a patient's condition stabilized, he/she was usually weaned off mechanical ventilation and extubated after the completion of the initial treatment.

Data sampling
The following data were collected: age, sex, Hunt and Kosnik (H&K) grade, treatment modality (coil or clip), number of PaCO 2 measurements, PaCO 2 levels, maximum and minimum PaCO 2 levels during the first 2 weeks in the ICU, serum lactate levels, use of mechanical ventilation, modified Rankin scale (mRS) score at hospital discharge, DCI rate, mechanical ventilation duration, ICU stay duration, hospital stay duration, and hospital mortality.

Outcome measures
The primary outcome was the association of abnormal PaCO 2 with unfavorable neurological outcomes assessed using the mRS at hospital discharge [21]. The mRS is a measure of global disability and comprises the following seven outcome categories: no symptoms at all, no significant disability, slight disability, moderate disability, moderately severe disability, severe disability, and death. All the patients with SAH were evaluated in real time using the mRS at hospital discharge, and the findings were mentioned in their medical records. The neurological outcome was considered unfavorable when the mRS score was 3-6 and was considered favorable when the mRS score was 0-2. The secondary outcome was the association of abnormal PaCO 2 with the presence of DCI.

Statistical analysis
The patients were divided into two groups according to their neurological outcomes (unfavorable and favorable outcome groups). Descriptive statistics are used to summarize the demographic factors and baseline characteristics. The groups were compared using Student's t test or the Mann-Whitney U test, as deemed appropriate. Fisher's exact test was used to make the categorical comparisons. Alterations in the PaCO 2 levels were examined separately for each clinical grade and SAH outcome. Univariate and multivariate analyses were performed to determine the independent predictors of unfavorable neurological outcomes. The covariates of age (> 65 years), sex (female), H&K grade, treatment modality (coil or clip), maximum lactate levels during the first 24 h [4,8,22], and maximum and minimum PaCO 2 levels during the first 2 weeks in the ICU were included in the multivariate analysis based on published reports [23,24]. The Mann-Whitney U test was used to estimate the receiver operating characteristic (ROC) curve and area under the curve (AUC). For each continuous variable, the cutoff value that had the best combination of sensitivity and specificity was identified. The patients were subdivided further into four groups according to the maximum and minimum PaCO 2 levels during the first 2 weeks in the ICU. The statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria); more precisely, it is a modified version of R commander designed to add statistical functions that are frequently used in biostatistics [25]. A two-sided p value < 0.05 was considered statistically significant in all the analyses. Missing data were not replaced or estimated.
No significant differences in median PaCO 2 levels were observed between the unfavorable and favorable outcome groups (p = 0.11); however, the unfavorable outcome group had significantly higher proportions of hypocapnia (PaCO 2 < 35 mmHg; p < 0.01) and hypercapnia (PaCO 2 > 45 mmHg; p = 0.02) among all the blood gas analyses during the first 2 weeks in the ICU. The episodes of hypocapnia were not intentional, and most involved an associated pH level of > 7.45 and not < 7.35. The relationships between the neurological outcomes of each SAH severity (H&K grades) and PaCO 2 are presented in Additional file 1: Table S1. Although there was no significant difference in the median of PaCO 2 or proportion of hypocapnia and hypercapnia across H&K grades, there were significant differences in minimum PaCO 2 levels and outcomes across H&K grades.
Alterations in PaCO 2 levels during the first 2 weeks in the ICU according to neurological outcomes As Fig. 2 shows, during the early stage of hospitalization (days 1-5), PaCO 2 levels were lower in the unfavorable outcome group than in the favorable outcome group. Subsequently, about 1 week after hospitalization (days 6-8), almost no differences in PaCO 2 levels were found between the groups. However, in the late stage of hospitalization (days 9-14), PaCO 2 levels were again lower in the unfavorable outcome group than in the favorable outcome group.
Alterations in PaCO 2 levels during the first 2 weeks in the ICU according to disease severity (H&K grade) As Fig. 3 shows, during the early stage of hospitalization (days 1-5), PaCO 2 levels were lower in patients with moderate-to-severe disease grade (H&K grades III-V) than in those with mild disease grade (H&K grades I-II). However, after day 6, almost no differences in PaCO 2 levels were found between the patient groups.

Predictors of unfavorable neurological outcomes in patients with SAH
On multiple regression analysis (

ROC curve analysis
ROC curves of unfavorable neurological outcomes according to the maximum and minimum PaCO 2 levels in the ICU were constructed. Table 3 presents the respective AUCs, sensitivities, and specificities for predicting unfavorable neurological outcomes. The optimal cutoff values for the maximum and minimum PaCO 2 levels were 48.3 and 30.2 mmHg, respectively, during the first 2 weeks in the ICU.

Association between PaCO 2 levels and unfavorable neurological outcomes
We found that 78.8% of patients with a maximum PaCO 2 of > 48.3 mmHg and a minimum PaCO 2 of < 30.2 mmHg had unfavorable neurological outcomes, whereas only 22.8% of patients who did not have either of these PaCO 2 levels had unfavorable neurological outcomes (Fig. 4).

Association between PaCO 2 levels and DCI
DCI was observed in 14.6% of patients. The median period from admission to the occurrence of DCI was 8 (range, 7-9) days. The presence of DCI was confirmed during the first 2 weeks in the ICU. Additional file 1: Table S2 presents comparisons of the clinical characteristics between patients with DCI and those without DCI. On univariate analysis, there were no significant differences in age, sex, treatment modality, median PaCO 2 levels, and proportions of hypocapnia and hypercapnia between patients with DCI and those without DCI, although the number of patients with DCI was relatively small. On the other hand, the maximum PaCO 2 level, SD of PaCO 2 , and ICU and hospital stay durations were significantly greater in patients with DCI than in those without DCI. In addition, DCI was associated with unfavorable neurological outcomes. On multiple regression analysis (Additional file 1: Table S3), DCI was significantly associated with the maximum PaCO 2 level (OR, 1.1; 95% CI, 1.02-1.19; p < 0.05).
Alterations in PaCO 2 levels during the first 2 weeks in the ICU according to the presence/absence of DCI As Fig. 5 shows, although the number of patients with DCI was relatively small, patients with DCI showed increased PaCO 2 variability during the first 2 weeks in the ICU when compared with the findings in patients without DCI.

Discussion
In the present study, we demonstrated that both the maximum and minimum PaCO 2 levels during ICU management were significantly associated with unfavorable neurological outcomes in patients with SAH. Moreover, we investigated the thresholds for both hypocapnia and hypercapnia to examine the target PaCO 2 level during ICU management in patients with SAH, which might help in further prospective studies with adjustment for factors, such as respiratory settings and sedation.

Association between hypocapnia and neurological outcomes
Several previous studies reported that hypocapnia was associated with poor neurological outcomes or DCI [14,15].
Williamson et al. reported that severe hypocapnia, which was defined as PaCO 2 ≤ 30 mmHg, was significantly associated with poor neurological outcomes (OR, 4.52; p < 0.01) [15], and this finding is consistent with our result. However, they reported that moderate hypocapnia, which was defined as PaCO 2 < 35 mmHg, was not independently associated with poor neurological outcomes (OR, 1.30; p = 0.58) [15]. Hypocapnia was previously shown to be associated with brain tissue hypoxia and poor outcomes in patients with TBI [26,27]. Therefore, the current guidelines do not support the use of prophylactic and induced hypocapnia (PaCO 2 < 30 mmHg) [28]. Consistent with these findings and guidelines, avoidance of hypocapnia, especially a PaCO 2 level of < 30 mmHg, is presumed to be desirable in the management of patients with SAH.

Association between hypercapnia and neurological outcomes
As shown in Fig. 2, on days 1-14, the 75th percentile of PaCO 2 levels appeared to not be significantly different between the unfavorable and favorable outcome groups. This result indicated that some patients with unfavorable outcomes had higher PaCO 2 levels with greater variation when compared with the findings in those with favorable outcomes. It remains unknown whether a hypercapnia strategy to avoid hypocapnia in patients with SAH is permitted. The association between hypercapnia and poor outcomes has been suggested to be secondary to hypercapnia-induced cerebral vasodilation, increased intracranial pressure, and decreased cerebral perfusion [13,29,30]. Alternatively, a previous report involving SAH patients with a poor clinical grade (Hunt/Hess III-V) investigated the association between controlled hypercapnia and CBF or brain tissue oxygen saturation (StiO 2 ) [10]. The report mentioned that although controlled hypercapnia of up to 60 mmHg reproducibly increased CBF and StiO 2 , the PaCO 2 reactivity of CBF, which was calculated by dividing the change in CBF by the change in PaCO 2 , was the highest at PaCO 2 levels between 40 and 50 mmHg and was slightly decreased at higher PaCO 2 levels. Impaired PaCO 2 reactivity has been reported to be frequent after SAH, particularly in patients with a poor clinical grade, and associated with DCI [31]. Considering our current data and the importance of PaCO 2 reactivity and possibly the cerebral steal phenomenon, hypercapnia of up to approximately 50 mmHg might be permitted as a therapeutic strategy. However, there are limited studies, and thus, further investigations are necessary.

Association between PaCO 2 levels and DCI
Several previous studies reported that hypocapnia was associated with DCI [14,15]. In our study, a small number of patients had DCI, and no significant association was noted between DCI and hypocapnia. According to our current data, hypercapnia or PaCO 2 variability might have been associated with DCI. However, further studies are necessary.

Mechanism of abnormal PaCO 2 levels
The mechanism of the association between abnormal PaCO 2 levels and neurological outcomes is unknown, and it was difficult to discuss about the causal relationship of abnormal PaCO 2 levels owing to the retrospective nature of our study as well as previous studies and the absence of adjustment for factors, such as respiratory settings and sedation [13]. However, severe brain injury following SAH itself might directly induce abnormal PaCO 2 levels [15]. In the present study, we found that 22.8% of patients (13 patients) with PaCO 2 levels within the cutoff range (30.2-48.3 mmHg) had unfavorable neurological outcomes. In these 13 patients, the mean age was relatively high (70.4 ± 11.4 years) and 69.2% (9 patients) had H&K grades III-V. We concluded that these factors, besides the abnormal PaCO 2 levels, could also affect the outcomes. Meanwhile, unfavorable neurological outcomes were noted in 78.8% of the patients with PaCO 2 levels outside the cutoff range (26 patients). For these patients, we also determined the duration during which their PaCO 2 levels were outside the range. ABG analyses represent PaCO 2 level at a certain time point; we assumed that PaCO 2 level among blood gas samples was linear. Although there was no significant difference among the neurological outcomes, the duration outside the cutoff range tended to be longer in the 26 patients with unfavorable neurological outcomes compared with the 7 patients with favorable neurological outcomes (33.8 ± 15.6 h vs. 25.1 ± 9.7 h, p = 0.17). Regarding this point, we concluded that a significant difference could have been noted if the number of patients was large. Accordingly, further detailed study regarding the influence of the duration during which the PaCO 2 levels are outside the range on neurological outcomes seems to be necessary.

Clinical implementation
The findings of the present study collectively suggested that both hypocapnia and hypercapnia were associated with unfavorable neurological outcomes after SAH. Our   [3,32,33]. Thus, the intentional control of PaCO 2 might be beneficial, especially in the first 72 h after hemorrhage when early brain injury occurs [8].

Limitations
The present study had several limitations. First, this was a retrospective observational study conducted at a single center, and thus, there was potential selection bias. Moreover, uncontrolled confounding factors may have been present. Second, the neurological outcomes of patients after discharge were not assessed. Third, the possible impact of differences in the patient population, such as those brought about by comorbidities, on neurological outcomes could not be definitively excluded. Fourth, the vital signs, sedation dose, oxygen dose, end-tidal CO 2 level, and specific mechanical ventilator settings, including the minute volume, fraction of inspired oxygen, and positive end-expiratory pressure during each ABG analysis, were not sufficiently considered. Therefore, it was difficult to assess the causal relationship of abnormal PaCO 2 levels and examine whether the PaCO 2 levels obtained in this study were intentionally controlled. Fifth, the study sample size was relatively small. Sixth, as this study had a small number of patients with DCI, the association between DCI and outcomes was not sufficiently analyzed. Finally, as the current study examined the limited association between abnormal PaCO 2 levels and neurological outcomes in patients with SAH, the effect of the regulation of abnormal PaCO 2 levels was not determined. Furthermore, as details of the causes of abnormal PaCO 2 levels, such as inadequate use of sedatives, antipyretic agents, or muscle relaxants, and the oxygen concentration or pressure support during mechanical ventilation were not obtained owing to the retrospective nature of this study, we could not make definitive suggestions on the management of PaCO 2 in a clinical setting. The results of our study should be regarded as hypothesis generating. Therefore, further prospective studies, including randomized controlled trials, are required to confirm the effects of abnormal PaCO 2 levels on neurological outcomes in patients with SAH.

Conclusions
We found that hypocapnia and hypercapnia during ICU management in patients with SAH were significantly associated with unfavorable neurological outcomes. Further prospective studies are required to validate our findings and explore their clinical implications. Our findings may provide a scientific rationale for these future prospective studies.

Additional file
Additional file 1: Table S1. Baseline characteristics of the study population across H&K grades. Table S2. Association between PaCO 2 levels and DCI (univariate analysis).