Perioperative Extracellular Brain Free-Water Changes for Older Adults Electing Total Knee Arthroplasty with General versus Spinal Anesthesia: A Pilot Study

Neuroimaging

All MRI acquisitions were performed using a 3T Siemens Verio (Siemens Medical Solutions, Malvern, PA, USA) with an 8-channel head coil. For each participant, two MRI acquisitions were acquired: a preoperative scan performed within 1 week before surgery and a second scan performed 48 h postoperatively. Each scan session included a diffusion acquisition as part of an imaging protocol.

Diffusion-weighted images were collected using a single-shot echo-planar acquisition with two acquisitions, with no diffusion weighting six-gradient directions at b = 100 s/mm², and 64 gradient directions at b = 1000 s/mm². The directions were defined using an electrostatic repulsion scheme [10]. Each volume was collected using a TR/TE of 17300/81 ms, an isotropic 2-mm resolution, 73 contiguous slices, and a 256-×256-inch in-plane field of view.

Diffusion-weighted images were first preprocessed using the eddy_correct command as part of the FMRIB Software Library package to correct for eddy current distortions. Further motion correction was performed using in-house code. The b vectors were also rotated following motion and eddy correction. Diffusion-weighted images were then reviewed using a free-water diffusion tensor imaging analysis that evaluated the volume fraction of freely diffusing water, as well as the anisotropic diffusion properties of hindered water in the brain parenchyma. For this experiment, the diffusivity of free-water was set to 3×10^–3 mm²/s corresponding to values of freely diffusing water at human body temperatures [1, 11]. This produces maps of fractional anisotropy and FW that can be used for subsequent analyses.

Diffusion analyses were performed using voxel-wise statistics of the diffusion data with tract-based spatial statistics as part of the FMRIB Software Library toolkit. First, each subject’s fractional anisotropy image was registered to a 1×1×1-mm³ space using the FMRIB58_FA image as a template. Next, the fractional anisotropy images were averaged across each group and used to derive a fractional anisotropy skeleton that represented the center of all white matter tracts common to each group. The analyses were then repeated for FW images using the script tbss_non_FA.

Preoperative T1-weighted images were acquired with the following parameters: TR: 2500 ms; TE: 3.77 ms; 176 sagittal 1 mm³ slices, 1 mm isotropic resolution; 256×256×176 matrix, 7/8 phase partial Fourier. Fluid-attenuated inversion recovery images were acquired with the following parameters: 176 contiguous slices, 1 mm³ voxels, TR/TE = 6000/395 ms. T1-weighted and fluid-attenuated inversion recovery images were used to calculate a general measure of “brain burden.” This variable included lateral ventricle volume in cubed millimeters, leukoaraiosis (white matter hyperintensities, which are a marker of cerebrovascular disease [12]) volume in cubed millimeters, and entorhinal cortex thickness in millimeters, with z scores calculated from the entire sample of surgical and non-surgical participants (total n = 131). This was calculated as previously reported [4] except that the composite was created with higher values indicating more “brain burden” (larger ventricles, greater leukoaraiosis volume, and thinner entorhinal cortex).

Participants

All general and non-surgery participants (full control) have been described previously [4]. The current analysis includes preliminary data assessing potential differences secondary to the surgical and anesthetic approaches. To analyze potential perioperative differences, multiple groups of participants were used: spinal anesthetic, general anesthetic, and non-surgical. We post hoc matched the five spinal anesthesia participants on an individual level with five participants who received general anesthesia and five non-surgery participants.

Matching was done on a case-by-case basis using a yoked procedure. We matched primarily based on 1) age, 2) education, 3) sex, with 4) ethnicity/race also considered. We secondarily matched on cognitive function from neuropsychological assessment (refer to Hardcastle et al. [13] for a description of the neuropsychological measures). We matched general anesthesia and control participants to their yoked spinal anesthesia counterpart to keep age within two years, education within one year, sex the same, and where possible, ethnicity/race the same. We previously showed perioperative changes in white matter FW were associated with preoperative cognition [4]. Therefore, as a secondary matching variable, participants were considered a match if the potential match had normative cognitive domain scores (declarative memory, working memory, processing speed) from neuropsychological assessment within 1 standard deviation of the yoked spinal participant. Participants were also given, but not matched using, the Montreal Cognitive Assessment [14]. To make additional inferences without sample size restrictions, the entire general (n = 61) and control (n = 65) groups were also analyzed using the approach described in the statistical analysis section.

Anesthesia protocols

Protocols were standardized, with participants receiving total knee arthroplasty with either general or spinal anesthesia.

General anesthesia. Participants received intravenous midazolam (1–4 mg) followed by continuous femoral nerve block and single-injection subgluteal sciatic nerve block. These blocks included (with 20 and 30 mL, respectively) 0.5% ropivacaine as a bolus injection, and the continuous femoral nerve block was continued with 0.2% ropivacaine at an infusion rate of 5 to 10 mL/h. No opioids were added. Propofol, fentanyl, and rocuronium were used for anesthesia induction and intubation. Patients were ventilated with an air/oxygen mixture to maintain an end tidal carbon dioxide at 35±5 mm and FiO₂ between 0.5 and 0.7; anesthesia was maintained with inhaled sevoflurane and intravenous fentanyl and rocuronium. Propofol boluses were administered as clinically needed for maintenance.

Statistical analysis

Paired two-tailed t-tests were created using FMRIB Software Library General Linear Modelling software. In the first analysis, the pre- and postoperative FW maps were analyzed for the full general and full control (non-surgical) groups. The second analysis involved only studying the pre- and postoperative FW maps for the spinal anesthesia group as well as the matched control and general groups. This second analysis was limited due to the limited size of the spinal anesthesia group (n = 5). For this comparison, subsets of general (n = 5) and control (n = 5) participants were collected that matched the spinal anesthesia group, as described previously. A paired two-tailed tract-based spatial statistics analysis was performed using randomise within each group, comparing the pre- and post-FW maps. Because of the limited size of the groups, a leave-one-out analysis was also performed, where tract-based spatial statistics were rerun using only four of the participants from each group to reject the notion that the results were driven by outliers. This was then repeated five times corresponding to removing one of the five participants from each group.

Based on the limited sample size, we created threshold points for change using a range of t values: t = 0 and the critical one-tailed t values for a sample size of 5 (1.533, p < 0.1; 2.132, p < 0.05; 2.776, p < 0.025). We used one-tailed values because based on our previous work with the larger sample demonstrated only group-level free-water increases and no decreases in white matter following surgery [4], we expected directional (i.e., increased) free-water changes. Increased and decreased free-water voxel counts were then determined for each critical value threshold and compared. Lastly, we created a ratio of free-water increase-to-decrease for each threshold value to assess patterns of change for each group. For ease of graphing, the ratio values were then log-transformed. We also conducted a post hoc Threshold-free Cluster Enhancement (TFCE) with 5000 non-parametric permutations to check if results survived multiple comparison correction.

We used one-way analysis of variance to test between-group differences in participant descriptive and demographic variables using SPSS v. 25 (IBM, Chicago, IL).

General, spinal, and non-surgery

Given the pilot nature of the study and limited sample size, we used different statistical thresholds with tract-based spatial statistics to assess within-group patterns in FW changes in brain white matter. Across thresholds, the three groups (general anesthesia, spinal anesthesia, and non-surgery) had similar numbers of white matter voxels with perioperative differences in FW (“perioperative” includes a pseudosurgery for the non-surgery group). While this could be partially attributable to noise in the data given the limited sample size (see “Limitations and Future Directions”), the patterns of changes were different between the surgery and non-surgery groups in both the limited and full samples. With the limited samples, the non-surgery group had relatively more voxels with FW decreases than increases. In contrast, the two anesthetic approach groups had more voxels with increased than decreased FW. In addition, the general group appeared to have the largest increase to decrease ratio, with the spinal anesthesia group having a smaller increase to decrease FW voxel ratio. Given the limited sample size, interpretation of spinal and general anesthesia results is done with caution.

General and non-surgery

For the larger sample including general anesthesia and control participants, which is a replication of our previous results [4], there were apparent differences in patterns of FW change. The control group, with stricter thresholding, showed little change in FW. What changes there were indicated general decreases in FW. Previously, we showed [4] that when correcting for multiple comparisons using a threshold-free cluster enhancement no white matter voxels significantly changed in the non-surgical control group. This indicates general stability of the FW metric over a 1-to-2-week period without surgery. Any interpretation of changes in FW in the non-surgery group in the present analysis where multiple comparison correction was not performed (to match the approach for the smaller samples where multiple comparison correction would be overly stringent) should be made with caution.

In contrast, the larger general group had many areas of significantly increased white matter extracellular FW. Again, this replicated our previous work assessing perioperative FW white matter changes with the same sample [4].

The analysis method of free-water imaging separates diffusion signals into tissue and extracellular components [1]. While FW may not be specific to neuroinflammation, extracellular FW increases can indicate glial change, including neuroinflammation [1–3]. If the FW increases indicate a reactive inflammatory response, this response could be harmful, neutral, or beneficial depending on the nature and extent of the response [15]. It is possible that reactive inflammation could interact with the presurgical level of chronic inflammation by increasing the baseline level of chronic inflammation. Chronic neuroinflammation has been linked with Alzheimer’s disease, Parkinson’s disease, and multiple sclerosis (see [15, 16] for review), which are typically accompanied by brain atrophy and difficulty on immediate memory tasks.

Of particular interest in the larger general anesthesia and spinal-matching subsample is the apparent concentration of increased FW in the frontal lobes, although we note that there were FW increases throughout the brain white matter. Frontal white matter carries connections to the rest of the brain but has large pathways to subcortical and parietal regions [17, 18]. Frontal-subcortical circuitry includes connections between dorsolateral and medial frontal regions, the caudate nucleus, and the thalamus [17]. The thalamus is important for response to anesthesia as are frontoparietal regions and networks [19–22]. While we did not test whether FW changes are directly linked to the anesthetic, it is not surprising that FW increases after major surgery under anesthesia particularly affect frontal white matter.

Frontal-subcortical and frontoparietal circuits are important for attention, working and immediate memory, among other functions [23–25]. Our previous work suggested a positive association between presurgical working memory/immediate memory and acute FW increases [4]. While we did not assess relationships between preoperative brain structure or pathology and FW changes in this study, previous research gives evidence that preoperative brain integrity (e.g., less atrophy, lower amounts of markers of cerebrovascular disease) and cognition before surgery under general anesthesia are risk factors for postoperative complications [4, 26–28]. It is possible that acute FW changes in frontal regions plays a role in postoperative complications, although this remains speculative.

It remains unknown if FW increases specifically indicate neuroinflammation, chronic neuroinflammation, or acceleration of neurodegeneration. Results suggest that the choice of anesthetic approach might result in different levels of acute FW changes, but the longer-term clinical implications of this finding in cognitively well adults is unknown. FW is a metric used to examine neuroinflammation in neurodegenerative disorders (e.g., Alzheimer’s disease, Parkinson’s disease; see [29–31]), but its value in the postoperative environment as an in vivo metric is still novel. Testing these hypotheses will require, at a minimum, larger sample sizes of individuals electing major surgical intervention under general or spinal anesthetic approaches.

Limitations and future directions

Several limitations need to be highlighted. First, the spinal anesthesia sample is small. The larger study was prospectively and carefully designed to study the effects of total knee arthroplasty under general anesthesia, but five participants elected spinal anesthesia. They were followed to generate pilot data. When comparing general anesthesia, spinal anesthesia, and non-surgery groups (n = 5 for each group), all had similar numbers of voxels where FW changed significantly. Importantly, it is likely these numbers of voxels match, in part, because of higher unexplained variance (noise) due to a limited sample. For example, comparing patterns in Fig. 2 (a and b), we see with larger sample sizes the control and general groups have dissimilar numbers of voxels where FW is different from baseline. However, the preliminary data must be interpreted with significant caution, because while the numbers of changed FW voxels were similar across the three groups the patterns of increased to decreased FW voxels were visibly different. Results did not survive a more robust TFCE analysis but that was expected given limited power. A larger sample will address this limitation. If possible, a randomized study would be most effective.

Another limitation is that while free-water is sensitive to inflammation, it is not specific to it [3]. Other etiologies (e.g., edema) could be contributing factors. Addressing this limitation would require other imaging modalities (e.g., spectroscopy) to confirm neuroinflammation. Cerebrospinal fluid markers of inflammation could also be collected to confirm the interpretation of FW as a marker of neuroinflammation, although outside of spinal anesthetic approaches such collection is invasive. Additionally, we did not assess if various intraoperative factors including blood pressure variability, total anesthetic amount, and respiratory rate, or other perioperative factors like body temperature correlate with brain FW changes.

Future directions include acquiring larger samples of older adults undergoing major surgery with spinal anesthesia. We previously showed individuals with lower preoperative brain integrity had less increase in white matter free-water [4]. These same techniques may be applied to compare microstructural changes in participants receiving total knee arthroplasty under general versus spinal anesthesia. While recent research suggests no differences in the clinical outcomes of mobility and delirium in older hip surgery patients undergoing general or spinal anesthesia [32, 33], our findings suggest there might be acute differences in brain response to the procedures. Future research using larger samples will need to assess whether the apparent differences are robust and clinically meaningful. Research also is needed to examine the interaction between severity of preoperative memory impairment and type of anesthesia response on acute and longer-term brain and cognitive changes.