Intracranial-to-Central Venous Pressure Gap Predicts the Responsiveness of ICP to PEEP in Patients with Traumatic Brain Injury

Background: Mechanical ventilation (MV) with positive end-expiratory pressure (PEEP) is commonly applied in patients with severe traumatic brain injury (sTBI). However, non-indicators to predict the influence of PEEP on intracranial pressure (ICP) prevent the optimal use of PEEP. As the central venous pressure (CVP) could act as an intermediary to transduce the pressure from PEEP to ICP, we set up a new indicators PICGap (representing the gap between the baseline ICP and baseline CVP). The aim of the study was to explore the relationship between PICGap and the ICP responsiveness to PEEP. Methods: Total 112 patients with sTBI undergoing MV were finally enrolled. ICP, CVP, cerebral perfusion pressure (CPP), static compliance of respiratory system (Cst), and end-tidal carbon dioxide pressure (PetCO2) were recorded at initial level of PEEP (3 cmH2O) and adjusted levels of PEEP (15 cmH2O). PICGap was calculated by baseline ICP - baseline CVP (when PEEP=3 cmH2O). The patients enrolled were classified into either an ICP responder group or a non-responder group based on whether the increment of ICP when PEEP adjustment from 3 cmH2O to 15 cmH2O was greater than or less than 20% of baseline ICP. Parameters recorded above were compared between two groups and the prediction of ICP responsiveness to PEEP adjustment were evaluated by receiver operating characteristic (ROC). Results: Responder group had lower PICGap, lower baseline ICP, and higher baseline CVP compared with non-responder group. ROC analysis suggested that PICGap could act as a strongest predictive indicator for the ICP responsiveness to PEEP (AUC = 0.957, 95% CI: 0.918 - 0.996, p <0.001) compared with baseline ICP and baseline CVP, with a favorable sensitivity of 95.24% (95% CI: 86.91% - 98.70%) and specificity of 87.6% (95% CI: 75.76% - 94.27%) when the cut off value of 2.5mmHg was

However, concerning of the influence of PEEP on intracranial pressure (ICP) has been an obstacle of the optimal use of PEEP for a long time [6]. The influence of PEEP on ICP was first mentioned in the later 1970s [7,8]. During last three decades, several studies explored the relationship between PEEP and ICP, but without consistent results. Shapiro et al demonstrate that application of PEEP in the 4-8 cm H 2 O range caused an increase of ICP (>10 mmHg) [7]. Flexman and colleagues also found that alveolar recruitment maneuvers increased subdural pressure and reduced cerebral perfusion pressure (CPP) during neurosurgery [9]. Recent study by Boone et al found that every centimeter H 2 O increase of PEEP contributed to a 0.31 mmHg increase in ICP [10], and concluded that PEEP might exert adverse effects on cerebral hemodynamics through impeding cerebral venous return and elevating ICP in patients with sTBI. However, other studies didn't find that moderate to high level of PEEP (8-25 cmH 2 O) affect ICP, CPP, and CBF of sTBI patients with normal ICP or intracranial hypertension, and even has favorable effects on improving brain tissue oxygen pressure and saturation [11][12][13][14]. The discordance of the results might relate to several factors: (1) the individual heterogeneity exists mainly involving severity and baseline ICP [15]; (2) the dose-effect relationship between PEEP and ICP hasn't been fully studied; and (3) it is unclear whether PEEP directly affects ICP or indirectly through an intermediate.
Until now, there is still non-indicators to predict the influence of PEEP on ICP. Theoretically increasing intrathoracic pressure by PEEP may hinder cerebral venous return and increase ICP when MV for patients with sTBI, and the relationship between CVP and PEEP has been clarified [16,17]. Therefore, we hypothesized, (1) CVP could act as a mediator between PEEP and ICP, (2) the effect of PEEP on ICP depends on the baseline of ICP and baseline CVP according to Starling resistor model ( Figure 1).
Herein, we defined a new indicator P IC Gap, which represents the difference value between baseline ICP and baseline CVP (at initial PEEP), and explored the association between P IC Gap and the ICP responsiveness to PEEP adjustment. Germany) were initially included. The patients whose hypoxemia (SpO 2 <90%) still could not be corrected through increasing the FiO 2 more than 60% in combination with suction and intensive airway management were eventually enrolled in the study.
Exclusion criteria included: brain death, age below 18 or over 80 years, pregnancy, hemodynamic instability [heart rate >120 bpm or CPP (calculated by MAP-ICP) <60 mmHg], pneumothorax, pulmonary bulla, and acute myocardial infarction (elevated cardiac troponin T more than 3 times the normal upper limit accompanied by the ST-T change) etc. Approval for study conduct was granted by the clinical research ethics committee (no. ZPYYLL-2016-12), and written consent was obtained from all participants' next of kin.

Design and measurement
The treatment program was implemented referring to the Guidelines for the Management of sTBI [18].
All patients were in supine position at 30 degrees head of bed elevation and deeply sedated (0.05 mg/kg loading dose, followed by continuous intravenous infusion of midazolam 0.05-0.3 mg/kg/h and sufentanil 0.2 μg/kg/h) to maintain the Richmond Agitation-Sedation Scale (RASS) score of −5 and, thus, to remove the interference of cough and other neuronal and confounding factors on ICP. The ventilator settings remained consistent for each enrolled patient. The tidal volume was adjusted and maintained at 8 mL/kg of predicted body weight and the plateau pressure was maintained below 30 cmH 2 O. Support pressure was maintained at 12-14 cmH 2 O, initial PEEP was set at 3 cmH 2 O, and fraction of inspired oxygen (FiO 2 ) was set at 35% -50% to maintain pulse oxygen saturation (SpO 2 )>90%. The ICP was continuously monitored (Codman ICP ExpressTM, Johnson, USA) through an intraparenchimal transducers or ventricular catheter (Codman ICP Transducer, Johnson, USA) that was 5 associated with a closed external ventricular drain if it existed during each measurement. Both central venous and arterial catheters were inserted to measure intra-arterial MAP and CVP. CPP was maintained more than 60-65mmHg. The static compliance of respiratory system (Cst) recorded from ventilator was indexed to the predicted body weight of the patients. During the study, the end-tidal carbon dioxide pressure (PetCO 2 ) (monitoring by Drager Mainstream CO 2 device, SN: ASHM-0552, Dräger, Germany) was maintained at 30-35 mmHg by adjusting tidal volume and respiratory rate, in order to avoid any effect of CO 2 on ICP [19].
The stepwise increase of PEEP was set according to the method by Lim et al [20] when the hypoxemia persisted. Briefly, 100%-FiO 2 was set up and PEEP was increased stepwise (from 3 cmH 2 O to 10 cmH 2 O, and to 15 cmH 2 O) every 2 min, which was a recruitment maneuver known as "extended sigh". ICP, CVP, Cst, PetCO 2 , and CPP at the two levels of PEEP (3 cmH 2 O and 15 cmH 2 O) were measured respectively. After PEEP at 15 cmH 2 O maintained for 2 min, baseline ventilator setting was resumed.
Based on our research hypothesis and specific relationships between CVP and PEEP [16,17], P IC Gap and other measurements (Cst, CPP, ICP, and PetCO 2 ) were compared between two groups, and the prediction of ICP responsiveness to PEEP was tested by calculating the area under curves (AUC) of the receiver operating characteristic (ROC). Because there is no specified definition of the ICP responsiveness to PEEP adjustment, we stipulated that responder and non-responder referred as to an increment greater than or less than 20% of baseline ICP respectively when PEEP set up to 15 cmH 2 O.

Statistical analysis
Categorical variables are presented as numbers and percentages and were analyzed by Fisher's exact test. Continuous covariates, including hemodynamic variables ICP, CVP, CPP, Cst, PetCO 2 and CPP, were expressed as means ± standard errors. One-way analysis of variance (ANOVA) followed by the Bonferroni post hoc test was used for multiple comparisons. The predictive role of P IC Gap other related parameters recorded for ICP responsiveness to PEEP were tested by calculating the AUC of the ROC for ICP over the baseline value at the two levels of PEEP (3 and 15cmH 2 O). A p-value less than 0.05 was considered statistically significant. All statistical analyses were performed by using SPSS 20.0 for windows (IBM Co. NY, USA).

Results
From May 2016 to May 2019, a total of 112 patients were entered into the final analysis. Baseline characteristics of the study population between responder group (n=49) and non-responder group (n=63) were shown in Table 1

1.
The effects of PEEP adjustment on CVP and ICP between responder group and nonresponder group.
Adjustment of PEEP from 3 to 15 cmH 2 O increased the levels of CVP significantly in two groups ( Figure 1A). There were no significantly difference in increment of CVP (ΔCVP) between responder group and non-responder group (4.39±1.30 versus 4.25±1.58, p=0.174) ( Figure 1B) Figure 1C). 2. The predictive role of P IC Gap, baseline ICP, and 7 baseline CVP on the responsiveness of ICP to PEEP adjustment .
As showed in Table 1, P IC Gap, baseline ICP, and baseline CVP were significantly different between responder group and non-responder group, and no significant difference was found in the other variables. The predictive ability of P IC Gap, baseline ICP, and baseline CVP were test through ROC. As showed in Figure 3

Discussion
In the present study, based on our hypothesis that CVP was an intermediary which deliveries pressure from PEEP exerted to ICP, we found that ICP was increased after PEEP only when baseline ICP was close to CVP, i.e. P IC Gap was narrower in responder group than non-responder group (1.63±1.33 versus 6.55±2.46 mmHg) which mean the same increment of CVP (4.39±1.30 versus 4.25±1.58) could disappear the P IC Gap for responder group, but not for non-responder group (Figure 1). We also evaluated the possibility whether the P IC Gap, baseline ICP, and baseline CVP could predict the ICP responsiveness to PEEP in patients with sTBI. The results suggested that P IC Gap should be the strongest predictive indicator among the three parameters. The P IC Gap less than 2.5 mmHg could predict the ICP responsiveness to PEEP tuned up to 15 cmH 2 O. To our best knowledge, this is the first study to demonstrate that PEEP-induced changes of ICP depended on P IC Gap rather than PEEP itself.
Although the cerebral hemodynamic is not governed entirely by the extradural venous pressure due to normal ICP (8-13 mmHg) was higher than the venous pressure outside the dura (0-5 mmHg), the changes of extradural venous pressure transfer to the brain circulation might rely on the certain situation [22]. The degree of subdural venous collapse was related to the difference between ICP and extradural venous pressure, and this passive collapse acts as a variable venous outflow resistance.
Alteration of extradural venous pressure causes up-or downregulation of venous outflow resistance through the self-regulation of the degree of passive collapse according to the Starling resistor model [23].
CVP could act as a surrogate marker of extradural venous pressure, because the pressure falls in jugular venous was negligible when at supine position. The preliminary experiment also showed that the values of CVP were the same as that of jugular bulb pressure. According to the Starling resistor model [23], once the value of CVP after PEEP exceeded baseline ICP, venous outflow resistance would be down-regulated to the lower limit. In such a situation, the brain circulation would be impeded and ICP rise accordingly.

Relationship between PEEP and CVP has been validated by previous studies. Stepwise PEEP elevation
induces an increase of CVP [17]. An increase of 12 cmH 2 O of PEEP caused a more than 4 mmHg rise of CVP in current study, which was consistent with previous findings [17]. Thus, it was reasonable to infer that PEEP led to the elevation of CVP directly, and whether CVP after PEEP could increase ICP depended on the extent of CVP narrowing the P IC Gap.
The lower value of P IC Gap means that the CVP after PEEP is easier to exceed baseline ICP, then the Starling resistor would lose effectiveness as a result of the elimination of venous outflow resistance.
As indicated in Table 1, patients with responsiveness to PEEP adjustment had the relatively lower P IC Gap compared with the non-responder group. Thus, based on the hypothesis that CVP is an intermediary which connects PEEP to ICP, we found that PICGap, as a new indicator, could provide a rational explanation on the underlying mechanism, which also accounted for the individual heterogeneity proposed by Yang and colleagues [15].
Brain compliance is unfavorable in patients with sTBI because of the cerebral edema caused by injury.
In this case, cerebral venous return impeded by elevated CVP after PEEP would contribute to the ICP increasing after P IC Gap got narrowed till to zero. A study by Robba and colleagues investigated the effects of pneumoperitoneum and Trendelenburg position on ICP in non-brain injured patients (lower ICP) and demonstrated that pneumoperitoneum and the Trendelenburg position increased ICP [24].
There was no significant change in arterial blood pressure and CPP in the study. Although the CVP was not monitored in their studies, increased ICP might be due to obstruction of cerebral venous return theoretically [25].
Several studies used baseline ICP to predict the responsiveness of ICP to PEEP, and found that patients with lower baseline ICP had positive response to various PEEP [26,27]. There results were consistent with our findings. Those with higher mean baseline ICP experienced no significant changes of ICP during the alteration of PEEP. However, these studies have not clarified that the certain ICP value could predict the responsiveness of ICP to PEEP. Our results also showed that baseline ICP could not be a favorable predictive indicator compared with P IC Gap.
It should be mentioned that the responsiveness of ICP to PEEP may be influenced by compliance of respiratory system [28,29]. Patients with low-compliance lungs showed that cerebral hemodynamics and ICP were not influenced by the application of PEEP, because less compliance may not transmit the increased pressure to the entire intrathoracic space effectively. In current study, all of the enrolled patients had normal compliance (Table 1).
Several limitations of this study should be mentioned that the sample size is relatively small for a clinical study, and the impact of PEEP on the CBF has not been evaluated. However, we kept PetCO 2 maintaining at the normal level, and elevation of ICP was in a permissible range, thus we speculated that the CBF would be stable. Furthermore, although the P IC Gap is a dynamic marker, the P IC Gap couldn't change dramatically in the early stage of TBI for certain patient after neurosurgery, which ensured the predictive value individually.

Conclusions
The impact of PEEP on ICP depends on the GAP between baseline ICP and CVP. The P IC Gap could be a potential predictor for ICP responsiveness to PEEP adjustment in patients with sTBI.

Ethics approval and consent to participate
The study protocol was approved by the clinical research ethics committee of Shanghai University of Medicine & Health Sciences Affiliated Zhoupu Hospital (ZPYYLL-2016-12). Written consent was obtained from all participants' next of kin because enrolled patients in the study were in a coma state.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the present study are available from the corresponding author on reasonable request. Effects of PEEP adjustment on CVP and ICP between responder group and non-responder group. Adjustment of PEEP from 3 to 15 cmH2O increased the levels of CVP significantly in two groups ( Figure 1A). There were no significantly difference in increment of CVP (ΔCVP) 20 between responder group and non-responder group ( Figure 1B). A significant increasing of ICP was observed in responder group when PEEP tuned up from 3cmH2O to 15 mmHg and no change in non-responder group ( Figure 1C) PEEP, positive end-expiratory pressure; ICP, intracranial pressure; CVP, central venous pressure.

Figure 3
The predictive role of PICGap, baseline ICP, and baseline CVP for ICP responsiveness to