The Effect of Propofol Versus Isoflurane Anesthesia on Human Cerebrospinal Fluid Markers of Alzheimer’s Disease: Results of a Randomized Trial

INTRODUCTION

The most prevalent type of dementia is Alzheimer’s disease (AD). In AD, amyloid-β (Aβ) monomers oligomerize into insoluble plaques (thus decreasing monomeric cerebrospinal fluid (CSF) Aβ levels), and the soluble microtubule stabilizing protein tau deposits into insoluble aggregates (i.e., “tau tangles”) [1]. The clinical onset of AD dementia is preceded by gradual increases in CSF total tau levels and gradual decreases in CSF Aβ levels that occur over decades [1], which together produce an increase in the CSF tau/Aβ ratio [2]. However, it is unclear what factors initiate or promote these pathologic processes. A small percentage of AD cases are due to dominant genetic mutations, while up to 50% of AD cases may be caused by environmental exposures [3]. Preclinical studies suggest that anesthesia and/or surgery could be environmental exposures that increase AD risk by causing increased Aβ levels [4], Aβ oligomerization [5], increased tau levels [6, 7],tau phosphorylation [8], neuroinflammation [9], and neuroapoptosis [10] (reviewed in [11]). If anesthesia and/or surgery accelerate AD pathogenesis in patients, this could potentially contribute to delirium, postoperative cognitive dysfunction, or longer-term cognitive decline, such as the eventual development of AD. Two small human studies have shown that CSF total tau levels increase after anesthesia and surgery [6, 7], demonstrating that some of these preclinical findings translate to humans.

Laboratory studies also suggest that inhaled anesthetic agents may promote AD pathogenesis to a greater extent than propofol [12, 13]. In fact, propofolblocks isoflurane-induced apoptosis and Aβ oligomerization in cell culture and in the mouse brain [12, 13]. If these findings extend to humans, they could be clinically important because specific anesthetic agents could be utilized to help decrease perioperative effects on AD biomarkers and pathogenesis, particularly in patients at higher risk of developing AD. Tang et al. examined CSF AD biomarkers in 11 patients who received sevoflurane or propofol, but this trial was non-randomized [7] and did not examine CSF AD biomarker differences between anesthetic groups. To date, no human randomized clinical trial has examined whether propofol versus isoflurane have differential effects on human CSF AD biomarkers after surgery. Thus, we examined this question in a pilot randomized trial. We chose the CSF tau/Aβ ratio between patients treated with isoflurane versus propofol as our primary outcome measure, since the CSF tau/Aβ ratio has high sensitivity and specificity for the diagnosis of AD [2]. Because genetic factors (such as ApoE4 [14]) and vascular disease (such as hypertension, diabetes and hypercholesterolemia [3]) contribute to AD risk and to changes in AD biomarkers, we also examined whether genetic and/or intraoperative parameters would predict the extent of postoperative AD biomarker changes.

MATERIALS AND METHODS

RESULTS

A schematic of study enrollment is shown in Fig. 1. Approximately 80% (i.e., 104/129) of the patients we approached (all of whom were scheduled for cases with lumbar CSF drain placement) agreed to participate in the study. 41 of the 104 patients who consented could not be enrolled because a lumbar CSF drain was not placed (either because the surgeons decided it was unnecessary, or were unable to place it) or did not work properly, resulting in an inability to obtain CSF. We initially planned to draw CSF samples and measure AD biomarkers at 0 and 3 h. However, an interim analysis of 0 and 3 h samples from the first 19 patients showed no significant effect of time, treatment, or interaction between time and treatment for any of the AD biomarkers measured (p>0.05 for all comparisons; Supplementary Figure 1). These data are consistent with an observational time-course study that showed no change in AD markers until 10–24 h after the induction of anesthesia [7]. Based on these findings, we modified our IRB protocol to collect CSF samples at the time of lumbar drain placement and 10 and 24 h later. We did not include data from the patients with 0 and 3 h samples in the final analysis.

Baseline and intraoperative characteristics of the study patients who had samples obtained at 0, 10, and 24 h are presented in Table 1. None of the patients had a diagnosed neurodegenerative disease or MCI. To assess any potential effect of the extent of intracranial surgery itself on CSF AD biomarkers, we separated the surgical cases into 3 groups: 1 = peripheral neurosurgery without deep intracranial work (e.g., trigeminal nerve decompressions or CSF leak repairs), 2 = deep intracranial surgery (e.g., cerebellopontine angle tumor or acoustic neuroma resections), or 3 = miscellaneous cases (e.g., cortical surface meningioma resections). There was no difference in the surgical case mixture between patients in the isoflurane versus propofol groups (Table 1). No statistically significant differences were found between groups for any of the baseline variables except for total propofol dosage, which was higher (as expected) in the propofol group.

There was no effect of anesthetic group on the CSF tau/Aβ ratio (p = 1.000; our primary outcome, Fig. 2A). An effect of time on the CSF tau/Aβ ratio was detected (p = 2.002×10^–6; Fig. 2A), but there was no interaction between time and anesthetic group on the CSF tau/Aβ ratio (p = 0.527). Effect sizes and β coefficients from this repeated measures ANOVA model are presented in Table 2. An effect of anesthetic group on CSF tau levels was not detected (p = 1.000). An effect of time on mean CSF tau levels was detected (p = 1.985×10^–6; Fig. 2B). There was no interaction between time and anesthetic group on CSF tau levels (p = 1.000).

There was no effect of anesthetic group on CSF Aβ levels (p = 1.000). An effect of time on CSF Aβ levels was not detected (p = 0.152; Fig. 2C), and there was no interaction between time and anesthetic group on CSF Aβ levels (p = 1.000). An effect of anesthetic group on CSF p-tau levels was not detected (p = 0.236). An effect of time on CSF p-tau levels was detected (p = 0.005; Fig. 2D). There was no interaction between time and anesthetic group on CSF p-tau levels (p = 1.000). There was no effect of anesthetic group on the p-tau/Aβ ratio (p = 0.295), and no effect of time on the CSF p-tau /Aβ ratio (p = 0.302; Fig. 2E). There was no interaction between time and anesthetic group on the p-tau/Aβ ratio (p = 1.000).

To identify factors that might account for inter-patient variation in the perioperative CSF tau increase at 24 h (Fig. 2B), we used rank-based tests to examine 24 h tau change. Of the factors examined in a univariate analysis (Table 3), only intraoperative dexamethasone dosage was associated with the degree of perioperative CSF tau increase at 24 h (p = 0.04, unadjusted for multiple comparisons). However, this exploratory post-hoc finding did not reach statistical significance after correction for multiple comparisons (p = 0.76).

Next, we examined whether there might be a genetic basis for the variation in perioperative CSF tau increases. APOE4 is a well-studied genetic risk factor for sporadic AD (reviewed in [21]), and polymorphisms in genes associated with inflammation (such as IL-6 and its receptor GP130) and mitochondrial dysfunction (such as TOMM40, an outer mitochondrial membrane 40 homolog) have also been implicated in AD [22 –25]. Many of these genetic variants also alter CSF AD biomarker levels [26, 27]. Thus, we genotyped patients for these and other related genetic variants, and examined whether these genetic variants were associated with the 24 h CSF tau increase. We were unable to obtain DNA from 1 patient. The remaining 38 patients were genotyped for 9 markers in 5 candidate genes: APOE, GP130, IL6, Tau, and TOMM40. All nine markers met Hardy-Weinberg Equilibrium (p > 0.05) and were tested for association with the extent of CSF tau elevation (Table 4). Equality of variance between carrier and non-carrier groups was similar for all markers except rs157581 in TOMM40 (p = 0.013). Thus, a 2-sample t test based on unequal variance was applied to rs157581. For the other 8 markers, we used a 2-sample t test with pooled variance. None of these genetic polymorphisms was associated with the 24 h CSF tau level increase (Table 4).

DISCUSSION

The data presented here show no difference in the increase in the CSF tau/Aβ ratio between patients undergoing neurosurgical/otolaryngologic procedures who were randomized to receive propofol or isoflurane. The CSF tau/Aβ ratio and tau levels increased over time, but these increases were anesthetic type independent. In vitro and animal model studies suggest that anesthetic drugs and surgical stress affect AD pathways (reviewed in [11] and [28]), and that inhaled anesthetics, such as isoflurane, may have a greater effect on these pathways than propofol [12 , 29]. However, the patient data presented here do not demonstrate a greater effect of isoflurane than propofol on AD pathways/markers.

Baseline and intraoperative characteristics were well matched between anesthetic groups (Table 1). Propofol dosage was the only characteristic that differed significantly between groups, as expected due to randomization. Although this is a pilot study, it is based on a relatively large data set- we obtained 165 perioperative CSF samples from 63 patients (Fig. 1). There are only four prior reports examining perioperative human CSF AD biomarker changes [6 , 31]; this study is larger than all but one of these [31]. This is also the only randomized study comparing the effects of propofol versus inhaled anesthetics on CSF AD biomarkers, a key question based on preclinical studies [12, 13] and the significant pharmacologic differences between propofol and isoflurane. Few perioperative CSF AD biomarker studies have been conducted despite the importance of understanding whether anesthesia and surgery modulate AD pathogenesis [11], because it is difficult to collect CSF samples from human surgical patients. This is the largest human perioperative CSF AD biomarker study performed using standardized laboratory measurements [32] and assays validated by the Alzheimer’s Association Global Biomarkers Consortium [33, 34]. This helps ensure the reproducibility of these results, a scientific goal promoted by the National Institute of Health [35]. We measured AD biomarkers in the ADNI biomarker core laboratory according to the ADNI protocol [32]; thus, the measurements reported here can be directly compared to those of the ADNI studies (as described below).

This study provides data about perioperative CSF AD biomarker variance and correlation between measures over time, thus allowing a post-hoc power analysis/sample size calculation for future studies on the biomarker measured here. Based on our data and our statistical model’s correlation coefficient between repeated measures (0.41), a future study with 18 propofol-treated patients and 21 isoflurane-treated patients with the same repeated measures design would have 80% power (with an α= 0.01) to detect a 20% difference in tau levels between groups. A 20% change in CSF tau levels would be clinically significant. In the ADNI study, a 20% increase would move the median CSF tau level found in normal individuals almost halfway toward the median value seen in patients with MCI [2]. This confirms that our study had adequate power for detecting a clinically significant tau difference between groups. To investigate differences in Aβ, p-tau, the tau/Aβ ratio, and the p-tau/Aβ ratio with these same statistical parameters (20% difference, 80% power, α= 0.01) and our observed means and correlations, a similar future study would require 45, 73, 72 and 16 patients per anesthetic group, respectively.

The mean tau value among patients at 24 h after surgery was 122 pg/ml (Fig. 2A), which exceeds the mean tau level in patients with MCI, and equals the mean tau level in AD patients in the ADNI study [2]. CSF tau levels > 93 pg/ml have a greater than 90% positive predictive value for AD [2]. We did not detect a difference in Aβ levels over time (Fig. 2C), although we cannot rule out the possibility that we were underpowered for this outcome as described above. Nonetheless, the tau/Aβ ratio (Fig. 2A) clearly increased over time due to the rise in CSF tau levels (Fig. 2B). Most earlier studies [6 , 36] but not all [31] have reported similar CSF tau and tau/Aβ ratio elevations after other surgery types.

In AD, CSF tau levels progressively rise prior to the onset of hippocampal atrophy, memory deficits, and clinical dementia [1]. Elevated CSF tau levels have also been detected in boxers [37, 38] and in individuals with traumatic brain injury [39, 40], and may predict long-term cognitive outcomes [39]. Nonetheless, the long-term trajectory of the acute 24 h postoperative changes in CSF tau levels reported here is unknown. It is also unknown whether acute postoperative CSF tau increases correlate with altered short- or long-term neurocognitive function. We are examining these questions in the MADCO-PC study Markers of Alzheimer’s Disease and neuroCognitive Outcomes after Perioperative Care https://www.clinicaltrials.gov/ct2/show/NCT01993836?term=Miles+Berger& rank=1), which will clarify whether anesthesia and surgery are associated with long term changes in AD biomarkers, and whether such changes are associated with long term cognitive changes.

Interestingly, in a secondary univariate analysis, intraoperative dexamethasone dosage was inversely correlated with the perioperative increase in CSF tau levels. This finding is from a secondary analysis and did not remain statistically significant after multiple comparison correction, but it provides hypothesis-generating preliminary evidence for future research. Glucocorticoids such as dexamethasone have pleiotropic effects on neural gene expression [41, 42], and increase tau gene expression in vitro. This suggests that dexamethasone does not decrease CSF tau levels simply by reducing tau gene expression. Instead, perioperative CSF tau elevations may occur via to a neuroinflammatory mechanism (which dexamethasone suppresses), a hypothesis that could be tested in a new trial. Although we found an inverse correlation between intraoperative dexamethasone dosage and the extent of perioperative CSF tau elevation in this secondary analysis, we found no correlation between several genetic polymorphisms that modulate the inflammatory response and the extent of perioperative CSF tau elevations. However, it is possible that we were underpowered to detect this genetic correlation (i.e., a type II error). Larger future studies will be necessary to examine a potential genetic basis for perioperative changes in CSF tau levels.

One caveat to these findings is that all patients underwent direct surgical manipulation of the dura and blood-brain barrier, and many underwent direct intracortical brain surgery. The effects of surgical manipulation of the dura and brain could outweigh the effects of anesthetic type on CSF AD biomarkers, which could explain why no anesthetic group difference was seen here. In this case, there could be differential effects of propofol versus isoflurane on CSF AD biomarkers after other types of surgery. We are currently examining this possibility as a secondary outcome in the MADCO-PC trial.

In conclusion, we found no difference in the CSF tau/Aβ ratio in patients who were randomized to isoflurane versus propofol anesthesia, although the lack of an anesthetic effect here may be due to a type II error. Future studies would need approximately 20–70 patients per group to have adequate power to detect anesthetic differences in these CSF AD biomarkers. Nonetheless, our data demonstrate that neurosurgery and otolaryngology procedures are associated with 3-fold increases in the CSF tau/Aβ ratio and total tau levels at 24 h, into the same ranges seen in AD [2]. The clinical and prognostic significance of these perioperative CSF tau/Aβ ratio and tau level increases is currently unknown and warrants further research.