Cerebral Autoregulation, CSF outflow resistance and outcome following CSF diversion in Normal Pressure Hydrocephalus

Introduction Normal pressure hydrocephalus (NPH) is not simply the result of a disturbance in cerebrospinal fluid (CSF) circulation, but often includes cardiovascular comorbidity and abnormalities within the cerebral mantle. In this study, we have examined the relationship between the global autoregulation pressure reactivity index (PRx), the profile of disturbed CSF circulation and pressure-volume compensation, and their possible effects on outcome after surgery. Materials and methods We studied a cohort of 131 patients, investigated for possible NPH. Parameters describing CSF compensation and circulation were calculated during the cerebrospinal fluid (CSF) infusion test and PRx was calculated from CSF pressure and arterial pressure recordings. A simple scale was used to mark the patients’ outcome 6 months after surgery (improvement, temporary improvement, and no improvement). Results PRx was negatively correlated with Rout (R=−0.18; p=0.044); patients with normal CSF circulation tended to have worse autoregulation. The correlation for patients who were surgically-managed (N=83) was: R=−0.28; p=0.03, and stronger in patients who improved after surgery (N=64; R=−0.36; p=0.03). In patients who did not improve, the correlation was not significantly different from zero (N= 19; R=0.17; p=0.15). There was a trend towards higher values for PRx in non-responders than in responders (PRx =0.16+/− 0.04 vs 0.09 +/−0.02 respectively; p=0.061), associated with higher MAP values (107.2+/−8.2 in non-responders vs 89.5+/−3.5 in responders; p=0.195). The product of MAP* (1+PRx), proposed as a measure of combined arterial hypertension and deranged autoregulation, showed a significant association with outcome (greater value in non-responders; p=0.013). Conclusion Autoregulation proves to associate with cerebrospinal fluid circulation, and appears strongest in shunt responders. Outcome following CSF diversion is possibly most favorable when CSF outflow resistance is increased and global cerebral autoregulation is intact, in combination with arterial normotension.


Introduction:
Normal pressure hydrocephalus (NPH) is a neurodegenerative disease characterized by enlarged ventricles, and Hakim's triad: gait disturbance, urinary incontinence, and dementia. Although prevalence estimates vary, it is known as the only "treatable dementia", and thus warrants ongoing research into its pathophysiology, early diagnosis, and treatment 1,2 . The pathophysiology of NPH, although better understood nowadays, remains to be elucidated. It can be divided in three components: disturbed CSF circulation, poor pressure-volume compensation and, possibly, abnormal Cerebral Blood Flow (CBF). Whether the regulatory mechanisms of CBF are disturbed cannot be concluded based on previous studies, primarily because almost every study has focused on demonstrating abnormal CBF patterns, which differ from our previous findings and current hypothesis on the importance of preserved autoregulation in outcome after shunting. Momjan et al (2004) showed reduced CBF in NPH patients white matter with a worsening pattern as the proximity to the ventricles increased. To the best of our knowledge, a few previous studies [Matthew et al, (1975 and1977), Schmidt et al (1996) and Tanaka et al 1997) have correlated normal CBF before shunting with a positive outcome. Those studies, however, only relied on measurements of CBF alone and there are a lot more that contradicted this finding or were inconclusive on whether higher or lower CBF was associated with a better post-surgical outcome [1][2][3][4][5] . Other studies have indicated that the ability of the brain to maintain a constant CBF in response to changes in cerebral perfusion pressure (CPP) -termed cerebral autoregulation, may be of importance in the pathophysiology of NPH 6,7,18 . In support of this, patient outcome after shunting has been linked with preserved autoregulation pre-shunting 16,22, ; however, the association was rather weak and could not be replicated in a further study 21 .
The role of the CSF dynamics, characterized by resistance to CSF outflow (R out ) and other pressurevolume compensatory parameters, is still ambiguous in NPH. Rout, calculated from routine CSF infusion tests, has been used as the most reliable infusion test -derived parameter in shunt response prediction. The road to the use of Rout as a diagnostic and outcome prediction tool was lead from early on with studies from Borgensen et al (1981 &1992) who reported a >80% post-shunting success rate for patients with Rout ≥ 13 mmHg*min/ml. The milestone dutch NPH trial 1 provided the necessary evidence for the use of the most common threshold for Rout, 18 mmHg*min/ml, reporting a 92% success rate. This trial pioneered in quantitavely reporting clinical investigations for the patients before and after shunting, however those highly positive results could not be replicated ever since.
The current knowledge on Rout and the different proposed thresholds for diagnosis in NPH are summarized on Table 1.
We previously reported a reciprocal relationship between the transcranial Doppler-based autoregulation index Mx and R out in non-shunted patients under investigation for NPH, such that those with lower R out , had worse cerebral autoregulation 11 . We hypothesized that in patients with low R out (suggesting normal CSF circulation), disturbed cerebral autoregulation may be in part responsible for the clinical symptoms. Similarly, in a cohort of both non-shunted and shunted patients, we found a reciprocal relationship between an intracranial pressure (ICP)-based marker of cerebral autoregulation (the cerebral pressure reactivity index-PRx) and R out 10 . Although these previous investigations direct us towards a negative correlation between R out and brain autoregulation in NPH patients, the external validity of this relationship is limited by the relatively small sample sizes, and the relevance of the relationship with respect to patient outcome is unclear.
The objectives for this study were first to confirm the relationship between cerebrovascular functionas measured by the pressure reactivity index (PRx), and CSF dynamics and circulation (characterized by R out ), and secondarily to examine how autoregulation (PRx) and systemic vascular dysfunction may relate to patient outcome following surgical management.

Data acquisition
Continuous arterial blood pressure (ABP) was measured using a Finapress finger cuff (Ohmeda, Englewood, CO), 11 , while continuous waveforms of ICP (as indicated by lumbar CSF pressure) were obtained by connecting the lumbar puncture needle to a fluid filled pressure transducer (Edward Lifesciences TM ) and pressure amplifier (Spiegelberg or Philips). In all patients, adequate CSF pressure and arterial pulse waveforms were assured. All data and digital signals were recorded at 30-100 Hz and then processed using ICM+ ® version 8.3 (University of Cambridge Enterprise Ltd) 33,34 Signal processing Artifacts (movements, coughing, dislocations of finger blood pressure cuffs, etc.) were manually removed in every patient's raw data file prior to analysis. The pulse amplitude (AMP) of ICP was calculated as the fundamental amplitude in the frequency band of 40 to 140 beats per minute. The magnitude of slow waves of ICP was quantified as the square-root of the power of the signal in the frequency band of 0.3-4 cycles per minute 8,9,19 .
The pressure reactivity index (PRx) was calculated as the correlation coefficient between 30 consecutive 10-second averages of mean ICP and mean ABP (MAP). To eliminate, as accurately as possible, the interference of gradually increasing ICP on the slow waves of ICP implicated in the PRx calculation, ICP was first detrended using a moving average high-pass filter [ Figure 1]. The cerebrospinal compensatory index (RAP) was calculated similarly as the correlation between 30 consecutive 10-second averages of mean ICP and AMP 8,13,14 .
The aforementioned variables were averaged during the baseline and plateau phases of the infusion test as well as during the entire test (duration 25-40 minutes) [ Figure 1].
The resistance to CSF outflow was calculated using a simple model integrated in our ICM+ software 8,11 . The algorithm and the model used to calculate CSF resistance and elasticity has been reported in detail before and has been subject to modifications and improvements 8,11,13 . In some cases, strong vasogenic waves or elevation of the pressure above the safe limit of 40 mm Hg precludes precise measurement of the final pressure plateau. Computerized analysis produces results even in difficult cases in which the infusion is terminated prematurely. In order to avoid physiologically impossible and inaccurate measurement, only static R out was used for further analysis 34 . The slope of the ICP Amplitude -mean ICP line (AMP-P line) was determined by regression analysis of AMP and mean ICP using scatter plot charts provided as a function of ICM+ 8,33 .
The product MAP* (1+PRx), produces a variable indicative of magnified vascular hypertension when autoregulation is impaired and attenuated vascular problems when autoregulation is preserved (PRx<0). As a concept similarly used before for ICP 33 we propose this product as a method that reflect the combined burden of hypertension and dysautoregulation in NPH patients.

Patient follow-up
After having undergone lumbar infusion studies, patients were scheduled for follow-up in the hospital Hydrocephalus Clinic and a clinical decision on their further management was made. Patients and their families were carefully counseled and offered either a ventriculoperitoneal or ventriculoatrial shunt that included a programmable valve or an endoscopic IIIrd ventriculostomy (ETV). Outcome was determined from the hospital records. A simple scale, previously reported 25 , was used to classify patients who were shunted or underwent ETV, according to the clinical improvement of their symptoms after their operation. The scale classifies patient outcome as improved sustainably for 6 months (=1), improved for 3 months and then stopped improving or worsened (=2) and not improved after neither 3 nor 6 months (=3).
A subgroup analysis was performed to compare ETV patients with shunted patients based on both the studied parameters and the outcome after shunting. Chi-square test was used to compare the outcome in ETV vs shunted patients

Statistical Analysis
After checking the normal distribution of data, Pearson's correlation test was used to assess the correlation between the different parameters described above. Paired-samples t-tests were used to compare means and the difference between baseline and plateau measurements within the group.
Independent samples t-tests were used to compare the means and the difference between in-cohort, separately defined groups of patients. Chi-square was used for qualitative data. Significance level was set at 0.05. Non-parametric tests (sign-ranked and Kruskal-Wallis) were used to compare results between the different outcome after shunting groups. All statistics were performed using the R software, version 3.3.1.

Results
The mean age of the cohort 73 (+/-7) years and the male to female ratio was ~7:5 (77 males, 54 females). All of the patients except 8 had possible iNPH, without any clear etiology. Those 8 patients are the ones below the age of 55 and the NPH etiology was post-hemorrhagic, aqueductal stenosis, Chiari malformation (type II) or congenital hydrocephalus which had been treated in the past, had become shunt-independent and the patient had presented anew with symptoms resembling hydrocephalus.
CSF compensatory and autoregulatory parameters at baseline and during the infusion are presented in  (PRx<0)) are given in Table 3. Increase in amplitude of ICP pulsation (dAMP) and compensatory reserve index RAP (dRAP) during the infusion studies were found to be significantly different between the autoregulating and non-autoregulating group.
Interaction between PRx, R out and other compensatory parameters R out was raised in 52 patients (>13 mm Hg*min/ml) and was within the 'normal' range (3.5 to 12.8 mm Hg*min/ml) in the remaining patients.  Shunted patients did not show a significantly more positive outcome than ETV patients (xhi-square = 0.0102; p-value= 0.91.

DISCUSSION
The relationship between the resistance to CSF outflow and pressure-reactivity index, although not very strong, was confirmed to be significant. Furthermore, an index of vascular function -as indicated by the combination of cerebral autoregulation and mean arterial pressure, was greater in patients with a lower chance of improving following shunt or ETV.
It is a very difficult endeavor to attempt utilizing two very conflicting aspects of NPH, Rout and autoregulation as a possible strong tool in NPH diagnosis and outcome prediction. It is unfortunate that after many years of very optimistic and polarly opposite, pessimistic views on both aspects there is uncertainty over those fundamental entities; the way forward from imaging and clinical scores appears inhibited. Etiology and pathophysiology of NPH need to be clarified as much as outcome prediction is in need of accuracy. Using ICP and autoregulation measurement techniques remains of the utmost importance in NPH until the achievement of optimal clinical management and outcomes; this includes many more studies to clarify and finally sort the ongoing discrepancies over definition, ambiguous diagnosis and usefulness of CSF dynamics parameters and role of the cerebrovascular bed 6,7,12 .
Tools such as infusion studies and ICM+ software have been the same since 2003 33,34 . We also have homogeneous criteria for admitting patients for hydrocephalus investigations in the CUH Hydrocephalus Clinic, introduced before 2003 and remaining unchanged. However, we cannot account in detail for the complete NPH assessment and referral process of our plethora of different neurosurgical consultants and this deviates from the initial purpose of our retrospective study.
High R out , despite criticism 1-5,10,25 , seems to be a major component of NPH, highlighting the importance of a disturbed CSF circulation 1,5,15 . However, when R out is used as a sole predictor of underwent extended lumbar drainage, as described above 12,16,17 .
Our results agree with the previous reported findings suggesting that autoregulation is better retained in patients with abnormally increased R out 10,11 . This perhaps paradoxical relationship between autoregulation and R out , now replicated with a greater power, was hypothesized to be due to the presence of cerebrovascular disease in those patients presenting with a normal or low R out 11,16,18,21 . We did not possess further data to support this hypothesis at that time. However, we have now observed that when 83 of the patients studied were clinically diagnosed with hydrocephalus and underwent surgery, the correlation between disturbed autoregulation and lower R out tended to be stronger. Even more interestingly, in the patients in whom the clinical decision was that they would most probably not benefit from surgery, no correlation between R out and PRx was present. This finding could possibly suggest a distinction between patients who might benefit from a CSF diversion and whose clinical symptoms could be caused by disturbed CSF circulation, with or without cerebrovascular disease, and at least not by vascular disease alone. Autoregulation appears to have the potential to serve as a supplementary tool for clinical decision-making. The presence of vascular disease, both as differential and as comorbidity in hydrocephalus, constitutes one of the biggest hurdles in clinical practice and management 17,22,23,28 . There is still no alternative clinical test and consensus to aid the clinical decision whether there is concomitant disturbance of the CSF circulation and cerebrovascular disease, or whether the CSF circulation is normal and the problem lies purely in the vasculature 10,22,24 .
PRx did not show any correlation with age, which is surprising, as significant correlation may be seen in anesthetized and ventilated patients after traumatic brain injury 9,20 . An additional similar correlation was shown between age and dysautoregulation in patients undergoing non-neurosurgical elective surgery. It appears that the age-autoregulation relationship (higher age associated with dysautoregulation) may be magnified by general anesthesia. On the other hand, R out is known to positively correlate with age 8,15 . A strong relationship was successfully replicated in our cohort of NPH patients; a possible investigation of different Rout thresholds for different ages remains to be performed separately, for better future directions. Sex does not appear to have any particular influence on R out .
The CSF infusion test seems to provide ideal conditions for measurement of cerebral autoregulation using different methods. We demonstrated a significant decrease in CPP during infusion. 8,13 . Another autoregulation-associated parameter, the magnitude of slow waves, correlated weakly but significantly with PRx. Slow waves, in context of measurement of PRx or transcranial-Doppler investigation, are carriers of the information about autoregulation mechanism. This correlation has not been reported before and it is worthwhile to consider its role in NPH patients, since both the frequency and the magnitude of slow waves are involved in possibly distinguishing NPH patients without vascular problems and predicting their benefit from a shunt insertion [26][27][28] .
Furthermore, the correlation between R out and PRx was found to be stronger in the patients that improved after shunting, and tended to be absent in the patients with no improvement. Outcome after shunting in NPH is very much multifactorial, and connecting it with one measured parameter is probably naïve 28,-31 . This can be demonstrated by the finding that neither PRx nor R out differed between the outcome groups but the correlation between PRx and R out was found to differ significantly. PRx is a global autoregulation index, which means that it reflects the state of autoregulation in whole of the brain.  11,35 .
Finally, it is important to point out that we presented a correlation between a lack of response to shunting and vascular (greater arterial pressure) as well as autoregulatory dysfunction combined.
There is an abundance of data suggesting an interaction between hemodynamics and CSF circulation 6,7,[28][29][30][31]37 . Strengthening the significant but not quite ideal predictive value of R out could prove valuable in clinical practice and spare patients from undergoing difficult, repetitive diagnostic procedures to determine their further management.

Limitations
ABP was measured non-invasively using a Finapres finger cuff device, and while this has been shown to be accurate compared to invasive methods [17] its accuracy in this particular patient group is unclear. Furthermore, infusion studies have a limited duration and do not always allow optimal calculation of the pursued autoregulation indices which typically require long recordings. Longer recordings serve as more ideal examples but are not always possible and are dependent on patient cooperation.
We had not obtained any measurements of cerebral blood flow during the infusion studies. Also, PRx is an indirect index of cerebral autoregulation. As mentioned above, it appears that a significant proportion of the hydrocephalic patients studied have a "grey zone" PRx. More studies are needed to determine the limits of PRx for hydrocephalic patients and furthermore compare different methods of autoregulation assessment. Also, the MRI scans of the patients available could provide us with some rough evidence of possible vascular disease, however they were not available for some of the patients in our cohort as MRIs for possible NPH have not always been performed routinely in our hospital. A possible avenue for future MR-based correlation of vascular disease would be to assess for the burden of periventricular and deep white matter hyperintensitites to explore this relationship.
The patients referred for infusion studies come from referrals from a multitude of neurosurgical consultants. Due to the retrospective nature of our study and to its design, it was impossible and not in our goals to possess all the clinical information regarding the detailed spectrum of the patients' disease, including the exact gait and memory scores. Based on evaluation from our experts, all of these patients were possible or probable NPH candidates undergoing further tests to ascertain the diagnosis and predict the outcome. Ventriculomegaly, gait impairment, dementia and urinary incontinence were diagnosed by specialists using the MMSE for cognitive assessment and gait scores from Tinetti Balance score and 10' walking test but we are unable to report those exact scores.
Finally, the sample size we had available for correlating with different outcomes is relatively small, especially for the patients with poor outcome. New studies with a stronger sample are required in order to validate our results related to patients' outcome.