Neural correlates of systemic lidocaine administration in healthy adults measured by functional MRI: a single arm open label study

Abstract

Introduction

Intravenous lidocaine is increasingly used as a nonopioid analgesic, but how it acts in the brain is incompletely understood. We conducted a functional MRI study of pain response, resting connectivity, and cognitive task performance in volunteers to elucidate the effects of lidocaine at the brain-systems level.

Methods

We enrolled 27 adults (age 22–55 yr) in this single-arm, open-label study. Pain response task and resting-state functional MRI scans at 3 T were obtained at baseline and then with a constant effect-site concentration of lidocaine. Electric nerve stimulation, titrated in advance to 7/10 intensity, was used for the pain task (five times every 10 s). Group-level differences in pain task-evoked responses (primary outcome, focused on the insula) and in resting connectivity were compared between baseline and lidocaine conditions, using adjusted P<0.05 to account for multiple comparisons. Pain ratings and performance on a brief battery of computer-based tasks were also recorded.

Results

Lidocaine infusion was associated with decreased pain-evoked responses in the insula (left: Z=3.6, P<0.001, right: Z=3.6, P=0.004) and other brain areas including the cingulate gyrus, thalamus, and primary sensory cortex. Resting-state connectivity showed significant diffuse reductions in both region-to-region and global connectivity measures with lidocaine. Small decreases in pain intensity and unpleasantness and worse memory performance were also seen with lidocaine.

Conclusions

Lidocaine was associated with broad reductions in functional MRI response to acute pain and modulated whole-brain functional connectivity, predominantly decreasing long-range connectivity. This was accompanied by small but significant decreases in pain perception and memory performance.

Clinical trial registration

NCT 05501600.

Editor's key points.

• •Intravenous lidocaine is used as a nonopioid analgesic intraoperatively and for chronic pain, but its actions on the brain are poorly understood.
• •In 27 adult volunteers, pain response task and resting-state functional MRI scans were obtained at baseline and then with a constant effect-site concentration of intravenous lidocaine.
• •Lidocaine reduced functional MRI responses to acute pain and modulated whole-brain functional connectivity, accompanied by small but significant decreases in pain perception and memory performance.
• •Lidocaine appears to modify pain by affecting both primary sensory processing and higher-level processing of noxious stimuli. This mechanistic understanding will inform further development of the clinical use of this important analgesic agent.

Intravenous lidocaine infusion is increasingly used in clinical practice to provide opioid-sparing pain relief, both in the perioperative setting1, 2, 3, 4 and for chronic pain.^5,6 Persistence of these effects for hours to days after termination of the infusion⁷ suggests involvement of higher-brain mechanisms, rather than direct sodium channel blockade in the periphery. There is evidence that lidocaine can improve postoperative cognitive function in high-risk patients,⁸ but cognitive effects during infusion have received little investigation. Empiric use of lidocaine in clinical practice varies widely; across 30 studies of postoperative lidocaine use, doses range from 1 to 5 mg kg⁻¹ h⁻¹ over periods from 1 h to 48 h after surgery.⁷ Lidocaine is particularly attractive as a perioperative nonopioid analgesic for patients with chronic pain or opioid tolerance, both particularly common in spine surgery, but conflicting efficacy has been shown in this population.9, 10, 11 A better understanding of the neural correlates of systemic lidocaine and its effects on brain function and cognition might help improve clinical use.

Little work has been done to examine the central effects of systemic lidocaine administration. One functional magnetic resonance imaging (fMRI) study in rats receiving lidocaine (but also anaesthetised with alpha-chloralose) showed no reduction in pain-related activation in somatosensory areas.¹² The only human neuroimaging study of systemic lidocaine used single photon emission computed tomography in nine female subjects under nonpainful conditions, showing increases in blood flow in the right anterior insula, right thalamus, and basal ganglia.¹³ Decreases in blood flow were only detected in the left posterior cingulate gyrus and precuneus.¹³

We conducted an fMRI study to determine the effects of systemic lidocaine on pain perception, cognitive function, brain response to acute pain stimulation, and changes in functional connectivity in humans. Our primary outcome was fMRI activation related to pain stimulation. We specifically hypothesised that pain task-related activation would decrease under lidocaine, specifically for changes in the insula cortex.

We also undertook secondary analyses to reveal changes in brain connectivity based on resting-state fMRI. In addition to a region-of-interest (ROI) to ROI analysis, we calculated atlas-free measures of local and global measures of connectivity change. We hypothesised widespread decreases in long-range functional connectivity, particularly between brain areas known to be involved in pain processing.

Additional secondary behavioural outcomes included ratings of pain perception and behavioural performance on cognitive tasks evaluating motor response speed, working memory, and long-term memory. Though effect sizes were expected to be small, we predicted that lidocaine infusion would show decreased pain ratings, slowed motor response, and impaired performance on both working memory and long-term recognition tasks.

Methods

Ethics and oversight

The study was approved by the local institutional review board (University of Pittsburgh, STUDY 21120115), and conducted in compliance with all applicable standards for the responsible conduct of research on human subjects. Informed consent was documented from all participants before any research interventions. Though a nonrandomised nontherapeutic study, it was preregistered as a clinical trial (NCT 05501600). Drug administration and subject monitoring were performed by a physician anaesthesiologist (KMV). An independent medical monitor was appointed to review adverse events.

Participants

Adult volunteers were recruited from the community and paid US$150 for their participation in this single-visit study, lasting about 3 h. Individuals were recruited using flyers and our institutional research participant registry (https://pittplusme.org/). Those expressing interest were screened for exclusion criteria: pregnant or attempting to conceive, BMI >40 kg m⁻², sleep apnoea, chronic pain, neurologic or psychiatric disease, anxiety, depression, cardiac rhythm disturbance, claustrophobia, metal implants, history of reaction to lidocaine, or regularly taking antiepileptic, antidepressant, antipsychotic, anxiolytic, stimulant, sleep-promoting, analgesic, or illicit drugs. Guidelines for preanaesthetic oral intake¹⁴ were followed. A negative urine pregnancy test was confirmed for females. All underwent preanaesthetic evaluation and screening to verify MRI compatibility. An electrocardiogram was examined for conduction abnormalities. A 22-G intravenous catheter was inserted in the hand or arm.

Study design

This was a nonrandomised open-label study, with a baseline control period followed by administration of lidocaine (Supplementary Fig. S1). After entering the MRI suite, subjects were connected to a monitor for standard vital sign monitoring.¹⁵ Lactated Ringer's solution was infused at 100 ml h⁻¹ throughout. Subjects performed the first cognitive task battery, pain task, and then resting-state fMRI scan. Lidocaine was then infused; once predicted steady-state was reached, the pain task, resting-state fMRI scan, and cognitive battery were repeated for the lidocaine condition. E-Prime v. 3.0 (Psychology Software Tools, Pittsburgh, PA, USA) was used to present instructions, control task timing, and record subject response data during the experiment. We utilised MRI-compatible projection, audio, and response glove systems.

Task descriptions

Before entering the scanner, an electric nerve stimulator (EZ Stim II, LifeTech) was connected to two small electrodes applied to the lateral side of the left index finger. Starting at zero, current was slowly increased until the subject reported 7 out of 10 pain intensity using a verbal rating scale (0 no pain, and 10 worst possible pain). During the two pain task periods, five 10-s long simulations were delivered in a block design at the previously set current intensity level. Pain ratings were reconfirmed with two brief stimuli once in the scanner, and current adjusted up or down as needed such that all subjects began the experiment with the nerve stimulator set to a level that they rated as 7/10 intensity. Ratings of pain intensity and unpleasantness were obtained following each pain task, under baseline and lidocaine conditions.

Motor response time was assessed by an auditory vigilance task, in which subjects pressed a button with their right index finger as soon as possible after hearing a tone; 10 tones were played, ∼ every 6 s, with unequal rest periods between tones, adding unpredictability. The average response time, ignoring the first two, was used in the analysis.

A combined recognition task was employed using a series of images made from patterns of geometric shapes appearing every 5 s. Subjects identified by pressing a button whether the image was new or previously seen. As used by others,¹⁶ these images were presented in a three-back fashion, with about half of the images not matching their three-back prior. After >60 s, images seen only once were presented again, effectively testing long-term memory (as the working memory buffer would be unable to store information for this period of time). Performance on both the three-back and long-term memory tasks was calculated using signal detection metric dʹ,¹⁷ calculated as a ratio of Z-scores for the hits (correct identifications) to the false alarms (incorrect false-positive identifications). Values were calculated in Excel v. 2403 (Microsoft, Redmond, WA, USA) using the NORM.S.INV function.

Lidocaine administration

Lidocaine (from multiple vendors) was administered using a precalculated dosing scheme based on targeting an effect-site concentration of 1.5 μg ml⁻¹ using a pharmacokinetic model¹⁸ implemented using the freely-available STANPUMP-R software (https://stanpumpr.io/). Bolus dosing was administered as a 1% solution, with a strategy designed not to overshoot the target concentration. The lidocaine infusion was maintained using a 2 mg ml⁻¹ solution and a programmable-rate pump (Alaris Infusion system, Becton, Dickinson and Company, Franklin Lakes, NJ, USA) kept in the MRI control room with 18–20 feet of small-bore (low-volume) extension tubing extending from the pump to the subject with a dedicated infusion line preloaded with the dilute lidocaine solution. The lidocaine line and the background carrier infusion were run to a two-connection 3-inch small-calibre connection such that initiation of the infusion and any rate changes would be realised with minimal dead space. Venous blood sampling from another large-bore venous catheter site was attempted in several subjects with the intent to analyse blood lidocaine levels, but this process was ultimately aborted because of futility.

Magnetic resonance imaging

A Siemens Prisma 3 T scanner (Seimens Medical Solutions, Malvern, PA, USA) was used for image acquisition with a 32-channel head coil. A blood oxygen-level dependent gradient echo planar imaging sequence was used for all functional scans, with Echo Time (TE)=30 ms, Time to repeat (TR)=800 ms, flip angle (FA)=52 degrees, field of view (FOV)=220 mm, and spatial resolution of 2.1 mm isotropic. Whole-brain coverage was achieved with 72 contiguous slices, acquired in an interleaved pattern. Phase-encode direction was anterior to posterior; bandwidth was 2290 Hz Px⁻¹; echo spacing was 0.58 ms. Performance gradient mode and advanced B₀ shim mode were used. Imaging acquisition acceleration was used with multi-band factor 8 and GRAPPA acceleration factor 2. A pair of spin echo field maps was obtained, temporally adjacent to each functional scan, using similar parameters in the same coordinate frame, except with TE=50 ms and TR=8160 ms. The pain task scan was 3 min long, resting-state scans were 8 min; these were obtained, in that order, both under baseline and lidocaine conditions. A high-resolution T1-weighted anatomical image was obtained (over 5.5 min) during drug titration, with a single-shot sagittal MP-RAGE acquisition, using TR=2.4 s, TE=2.15 ms, TI=1 s, FA=8 degrees, 1 mm isotropic resolution, FOV=256 mm, interleaved acquisition, and GRAPPA factor 2. All imaging data have been made freely available via the OpenNeuro platform (https://doi.org/10.18112/openneuro.ds005088.v1.0.0).

Behavioural data analysis

Statistical analyses of behavioural data were performed with SPSS Statistics 29.0.2.0 (IBM, Armonk, NY, USA). Data were analysed with descriptive statistics, and formal test for normality using the Kolmogorov–Smirnov test. Normally-distributed data was planned to be analysed by paired t-test, with pairs of data by subject comparing baseline and lidocaine conditions. All statistical analyses were two-tailed, with P<0.05 used as the threshold for significance.

Task fMRI processing and analysis

Functional MRI data were preprocessed using FSL (FMRIB, https://fsl.fmrib.ox.ac.uk/fsl/). A temporal filter cutoff of 20 s was set for high-pass filtering with a spatial smoothing step using a Gaussian kernel with full-width at half maximum of 4 mm. Non-brain tissue removal was done using Brain Extraction Tool (BET).¹⁹ Slight manual adjustments to the fractional intensity and gradient threshold parameters were done for each subject to achieve better extraction of the brain from non-brain areas. Functional data were corrected for B₀ inhomogeneity using the acquired field maps distortions using the topup function within FSL, which estimates susceptibility-induced off-resonance effects between the two fields maps acquired in opposite encoding directions.^20,21 Motion was calculated with MCFLIRT, and motion parameters and their temporal derivatives were included in the model. Motion outliers²² >1.5 mm root mean square were found in three scans and accounted for with modelling in the first-level analysis. Physiologic noise correction was performed with CompCor,²³ based on eigenseries calculated from temporal fluctuations in white matter and cerebrospinal fluid.

For the pain task, timing of the nerve stimulation was modelled using a gamma haemodynamic response function, including derivatives. In two subjects, the first pain stimulation did not occur due to technical error, and this was accounted for in task modelling. Group analysis used a paired mixed-effects model (FLAME stage 2). Resulting group average maps were thresholded for an adjusted P<0.05, after a cluster significance threshold of Z>2, correcting for multiple comparisons.

Functional connectivity processing and analysis

Functional connectivity analysis was performed on the resting-state data using Conn Toolbox.²⁴ Preprocessing included realignment (six-parameter, rigid-body transformation), susceptibility distortion correction (b-spline interpolation), slice timing correction, outlier detection, direct segmentation, smoothing, and Montreal Neurological Institute (MNI)-space normalisation. Potential outlier scans were identified as those with framewise displacement above 0.9 mm or global BOLD signal changes above five standard deviations. Denoising included CompCor²³ and regression of linear trends (two factors) within each functional run. Bandpass filtering was applied for 0.008–0.09 Hz.

ROI-to-ROI connectivity matrices were estimated using each pair of 132 regions in the Harvard-Oxford atlas (visual depiction shown on the Conn website, https://web.conn-toolbox.org/). ROIs were grouped using data-driven hierarchical clustering, based on both anatomical proximity and similar temporal patterns. Clustering reduces bias from (somewhat arbitrarily drawn) boundaries of atlas-defined ROIs and reduces the total number of statistical comparisons (increasing analysis power). Fisher-transformed bivariate correlation coefficients were estimated separately for each pair of ROIs. ROI-level inferences were based on parametric multivariate statistics, combining the connection-level random-effects statistics across all connections from each ROI. Correction for the number of connections set the family-wise false-discovery rate at P<0.05.²⁵

Local correlation (LCOR) maps estimate local coherence at each voxel, taking the weighted average of all short-range connections between a voxel and its neighbourhood (radius 25 mm full width at half maximum). Global correlation (GCOR) maps estimate the average of all short- and long-range connections between a voxel and the rest of the brain. LCOR and GCOR were computed from bivariate correlation coefficients between voxels, estimated using a singular value decomposition of the normalised signal time course with 64 components separately for each subject and condition. Voxel-level hypotheses were evaluated using multivariate parametric statistics with random-effects across subjects. Cluster-level results were thresholded using a voxel-level threshold of P<0.001 and a family-wise false-discovery rate set at P<0.05.

Results

Subject demographics and drug doses

We recruited 27 subjects (13 male), with mean age 31.4 yr, age range 20–55 yr. Subject characteristics are listed in Table 1. There were no adverse events. Brief and mild side effects (abnormal tastes, sounds, etc.) were almost universally experienced shortly after initial dosing, but did not persist for more than 5 min, and were always nonprogressive in nature. All subjects reported feeling no side effects at the end of the experiment. No subjects endorsed feeling restlessness or anxious during the scan. Total lidocaine dose was mean (sd) 1.8 (0.1) mg kg⁻¹, range 1.7–2.0 mg kg⁻¹. This was administered over 27.1 (2.5) min, range 21–33 min, with variation depending on technical issues prolonging the experimental period. Full details for each subject are in Supplementary Table S1.

Table 1.

Subject characteristics, nerve stimulation parameters, and pain ratings.

Subject	Sex	Age (yr)	Mass (kg)	Nerve stimulator intensity (mA)	Pain ratings after pain task
					Intensity		Unpleasantness
					Baseline	Lidocaine	Baseline	Lidocaine
1	F	23	83.2	8	7	6.5	6	5
2	M	55	82.5	11	5	5	4	5
3	M	24	72.7	18	7	7	7	7
4	F	20	53.9	9	7	4	7	5
5	M	29	75	17	7	6	6	5
6	M	44	97.1	12	7	5	5	5
7	F	54	57.2	7	7	6	7	6
8	F	20	78.3	7	7	7	4	3
9	F	26	79.4	12	7	7	7	4
10	M	28	86.1	8	7	7	2.5	1
11	M	36	113	12	7	6	7	6
12	M	40	109.1	15	6	6	6	6
13	F	40	64	9	7	7	7	7
14	F	26	56.8	13	6	6	10	8
15	M	32	52.3	24	7	5.5	6	5.5
16	M	29	81.9	6	8	7	8	7
17	F	25	68.2	14	6	5	7	6
18	M	37	99.8	8	6.5	6.5	7	6.5
19	F	21	67.5	7	7	7	7	7
20	F	29	87	22	8	7	7	5
21	F	33	106	5	5.5	7	5	6
22	F	20	58.1	9	6	8	6	8
23	M	21	74.8	26	6.5	5	7	6
24	F	27	51.2	5	7	8.5	7	8.5
25	F	48	59.1	8	9	7	8	7
26	M	42	105.5	19	7	7	7	7
27	M	20	73	13	6.5	6	3	3
Mean		31.4	77.5	12.0	6.8	6.4	6.3	5.8
sd		10.4	18.4	5.8	0.8	1.0	1.6	1.7
Minimum		20.0	51.2	5.0	5.0	4.0	2.5	1.0
Maximum		55.0	113.0	26.0	9.0	8.5	10.0	8.5

Behavioural results

Table 1 lists pain scores for each subject, obtained after the pain task under the no-drug baseline and lidocaine conditions. Average pain ratings are shown visually in Supplementary Figure S2. Ratings of both pain intensity and unpleasantness were normally distributed based on Kolmogorov–Smirnov testing. Nerve stimulator intensity values for each subject are listed in Table 1. Pain intensity scores (mean [sd]) at baseline (6.8 [0.8]) were slightly higher than with lidocaine (6.4 [1.0]), with a mean rating difference of 0.44 (95% CI 0.01 to 0.88, P=0.045). Pain unpleasantness scores at baseline (6.3 [1.6]) were higher than with lidocaine (5.7 [1.7]), with a mean rating difference of 0.56 (95% CI 0.11 to 1.0, P=0.016).

Motor vigilance task response times were not meaningfully different between conditions. The mean difference comparing lidocaine to baseline was ∼10 ms. Thus, no further analysis was performed on the motor response data.

Performance on the three-back task, which requires attention and executive function, was decreased with lidocaine infusion. The across-subjects distributions for dʹ values were normal, and thus a paired t-test was used for analysis. Recalling that dʹ is a Z-score-like measure, three-back performance at baseline (1.7 [0.8]) did not show a significant difference with lidocaine (1.5 [0.8]), with a mean dʹ difference of 0.24 (95% CI –0.04 to 0.51, P=0.087). These results are shown graphically in Supplementary Figure S3.

Recognition memory performance was decreased with lidocaine (Supplementary Fig. S3). Distributions for memory dʹ values were normal, and a paired t-test was used. Memory performance at baseline (2.2 [1.1]) was higher than with lidocaine (1.7 [1.0]), with a mean dʹ difference of 0.48 (95% CI 0.12–0.85, P=0.011).

Functional MRI responses to pain

Group average pain-related brain activation differences are shown in Figure 1. Consistent with our primary hypothesis, brain fMRI response to pain decreased with lidocaine in the insula bilaterally. Decreased pain responses were also seen throughout commonly described pain processing areas,26, 27, 28 including the anterior cingulate gyrus, mid-cingulate gyrus, left thalamus, bilateral primary somatosensory cortex, left cerebellum, bilateral putamen, right primary motor cortex, medial prefrontal cortex, and a small portion of the left hippocampus. Increases in brain response with lidocaine were seen in the parietal cortex bilaterally, suggesting modulation of activity in this association area. A list of all significant clusters of activation differences (baseline vs lidocaine) are included in Table 2. A more detailed list of local maxima within each cluster is in Supplementary Table S2, which describes in more granularity what are likely distinct areas of activity change that occurred near each other, forming a contiguous cluster. Overall, brain activity related to acute pain was decreased by lidocaine infusion in many areas throughout the neuromatrix commonly associated with processing incoming stimuli.²⁹

Fig 1.

Pain task functional MRI differences for baseline vs lidocaine condition averaged across subject. Cluster corrected (for multiple comparisons) thresholded at Z>2, P<0.05. Slice numbers refer to coordinates in the MNI-152 standard space template. L, left; M1, primary motor cortex; R, right; S1, primary somatosensory cortex.

Table 2.

Pain task functional MRI results for baseline vs lidocaine by cluster.

Cluster	Voxels	P	Z-max	Z-max coord			COM coord			Brain area(s) in cluster
				x	y	z	x	y	z
1	1366	<0.001	4.29	30	–20	66	40.1	–26	49.2	R primary motor and somatosensory cortices, R supramarginal gyrus, R superior parietal lobe
2	707	<0.001	3.59	–36	–12	–12	–35	–4.6	1.76	L hippocampus, L insula, L superior temporal gyrus, L putamen, L thalamus
3	516	<0.001	4.05	–10	60	10	2.78	54.6	14.7	Bilat medial prefrontal cortex, L anterior cingulate, R superior frontal gyrus,
4	508	<0.001	3.71	–50	–32	54	–53	–26	32.1	L primary somatosensory cortex, L supramarginal gyrus, L superior parietal lobe, L superior temporal gyrus
5	469	<0.001	4.06	4	2	36	1.24	3.89	33	Bilat mid-cingulate, corpus collosum
6	377	0.00368	3.57	32	8	4	36	–3.5	1.94	R insula, R superior temporal gyrus, R putamen
7	370	0.00425	3.4	–58	–56	2	–50	–54	12.9	L middle temporal gyrus, L angular gyrus
8	357	0.00554	3.69	–36	30	–4	–38	25.5	–6.2	L anterior insula, L inferior frontal gyrus
9	331	0.00953	4.32	–20	–52	–22	–21	–52	–25	L cerebellum
10	1884	<0.001	–4.97	8	–76	36	–3.9	–74	31.6	Bilat occipital lobe, Bilat cuneus, L angular gyrus,
11	548	<0.001	–4.49	42	–68	–2	36.3	–65	–9.9	R inferior temporal lobe, R occipital lobe, R fusiform gyrus, R lingual gyrus
12	471	<0.001	–4.07	30	–56	44	33.3	–57	42	R angular gyrus, R lateral superior parietal lobe

P, calculated P-value after correction for multiple comparisons; Z-max, maximum Z-score within cluster; Coord, coordinates in MNI standard space template; COM, centre of mass; R, right; L, left; Bilat, bilateral.

Functional connectivity results

Results regarding ROI-to-ROI connectivity are shown in the connectome ring in Figure 2 and in matrix form in Supplementary Figure S4. Connectivity was predominantly decreased with lidocaine. Hierarchical clustering yielded 17 groupings of ROIs with altered connectivity with lidocaine compared with baseline. A detailed list of connectivity differences and ROIs included in the clusters are tabulated in Supplementary Table S3. Figure 3 displays these connectivity changes anatomically distributed throughout the brain, both within and between hemispheres. There is a predominance of temporal lobe ROIs identified, but connectivity changes were also seen in frontal, occipital, parietal, and cerebellar areas, as well as deeper brain structures such as putamen, amygdala, hippocampus, and cingulate gyrus. For comparison with alternative methods, we show matrix connectivity differences for a nonclustered ROI-to-ROI analysis in Supplementary Figure S5, noting fewer connectivity changes identified by this more constrained analysis.

Fig 2.

Group differences in resting connectivity using hierarchical clustering with multivariate pattern analysis. Decreases in connectivity with lidocaine are indicated by warm colours according to the scale shown. a, anterior; AC, anterior cingulate gyrus; AG, angular gyrus; C, cerebellum; Forb, frontal orbital cortex; FusC, fusiform cortex; i, inferior; ICC, intracalcarine cortex; IFG, inferior frontal gyrus; ITG, inferior temporal gyrus; L, left; LOC, lateral occipital cortex; MidFG, middle frontal gyrus; MTG, middle temporal gyrus; OFusG, occipital fusiform gyrus; p, posterior; PaCiG, paracingulate gyrus; PaHC, parahippocampal cortex; PC, posterior cingulate gyrus; R, right; SCC, supracalcarine cortex; SMG, supramarginal gyrus; STG, superior temporal gyrus; TO, temporal occipital; V, vermis.

Fig 3.

Anatomical display of the location for significant region of interest (ROI)-to-ROI connectivity changes in the hierarchical clustered analysis also shown in Figure 2.

The effects of lidocaine on short- and long-range patterns of functional connectivity are further characterised in LCOR and GCOR results (Fig. 4). A detailed list of connectivity differences found in the LCOR and GCOR analyses is presented in Supplementary Table S4. LCOR analysis demonstrated a single area of decreased localised connectivity in the right inferomedial temporal lobe, extending into a portion of the cerebellum. GCOR analysis revealed seven independent clusters with decreases in long-range connectivity with lidocaine. To better visualise the data underlying the LCOR and GCOR analyses, example unthresholded maps are shown in Supplementary Figure S6, demonstrating that local correlations have quantitatively stronger coefficients compared with global correlations.

Fig 4.

Differences in local (top) and global (bottom) correlation analyses rendered on brain surfaces (left) and shown as transparent glass brain views (right).

Discussion

We observed significant effects of systemic lidocaine on brain function by fMRI. Despite only modest decreases in pain ratings and memory performance, significant changes in brain activity were observed. This occurred in the absence of observable sedation and with all participants subjectively feeling back to baseline while still receiving a lidocaine infusion, consistent with the relative safety of the dose used, noting that 2 mg kg⁻¹ was the maximum total dose our Institutional Review Board allowed after their review of the study protocol. Subsequent investigations at higher doses could potentially reveal more granular changes in cognitive task performance, and even more marked fMRI changes, but with a greater risk of neurotoxicity.

We centred our primary analysis and hypotheses on the insula given our prior evidence that this brain area discriminates between pain and rest,³⁰ and the elegant work by others showing the insula's intricate connectivity to other pain-processing brain areas.^31,32 The insula is also commonly held to sit at the intersection between the discriminative and affective dimensions of pain processing.³³ Our current results demonstrate robust differences in pain-evoked fMRI responses throughout the insular system, including the anterior insula (left, superior portion), mid-insula (bilaterally), and the posterior insula (right, inferior predominant). We also detected bilateral activation in the primary somatosensory cortex, which is not uncommon in pain studies, even with single-sided stimulation. Importantly, decreases in pain-evoked responses were seen bilaterally in primary sensory areas comparing lidocaine with baseline. Taken together, these results suggest that lidocaine affects both sensory discriminative and affective processing, which was reflected in reduced pain ratings for both intensity and unpleasantness. We note that anterior insula connectivity has correlated to anxiety,^34,35 and that anxiety can increase throughout an MRI scan using emotion-inducing tasks.³⁶ We have also previously demonstrated connectivity differences between experiencing pain (using the same nerve stimulation) and resting-state that localised specifically to the mid-posterior insula.³⁰ Notably, our connectivity analysis was not parcellated to resolve subregions of the insula, so inferences between our pain tasks results and other work reporting insula subregion connectivity changes is speculative.

Given the paucity of data on the neural effects of systemic lidocaine, it was difficult to frame a priori expectations for how lidocaine would affect brain connectivity. The LCOR and GCOR analyses demonstrate, on a regional basis, how whole-brain connectivity is modulated by lidocaine. Unlike the ROI–ROI analysis, the LCOR and GCOR analyses are not constrained by anatomical (atlas-based) boundary definitions. These two complementary analyses provide a data-driven approach to visualise changes in patterns of coherence, without atlas-imposed anatomical boundaries inherent in an ROI–ROI approach.³⁷ The patterns of underlying unthresholded LCOR and GCOR data demonstrate that local correlations are much stronger than global correlations. Our results show that lidocaine was associated with more widespread decreases in GCOR compared with only one significant change in LCOR (isolated to the medial temporal lobe). From this, we infer that the stronger underlying localised correlations in brain connectivity persist with this dose of lidocaine.

To maximise ability to detect meaningful differences in ROI-based connectivity, we used data-driven hierarchical clustering for grouping ROIs, taking into account both location and temporal correlations. This analysis showed widespread bilateral decreases in connectivity between anatomically distant and functionally distinct ROIs. Differences in GCOR, and comparatively fewer changes in LCOR, corroborate that the delivered dose of lidocaine predominantly decreased long-range functional connectivity. Although summarising across analyses is inherently nonquantitative, we note commonality of temporal lobe involvement in both the clustered ROI and GCOR analyses. Further, the alternative nonclustered ROI analysis (which is more statistically constrained by multiple comparisons) still identified temporal lobe structures as having significant decreases in connectivity with lidocaine. Taken together, these results indicate that lidocaine has predominant connectivity effects in the temporal lobe compared with other brain areas at the dose given. Whether this phenomenon is specific to the drug, the dose, or a general susceptibility of temporal regions is an open question for future investigations.

Strengths and limitations

We incorporated several rigorous design features, including the within-subject design, using effect-site concentration modelling for lidocaine dosing, and state-of-the-art fMRI acquisition and analysis techniques. Another practical strength was the single-visit design, which was very efficient for participant recruiting and throughput. These strengths notwithstanding, the study is not without limitations. Firstly, the sequential ordering of conditions was nonrandomised, with lidocaine data collection always following the no-drug baseline. Because of residual drug effects, the only alternative would be to have a two-visit randomised design, which would inevitably introduce retention issues and unbalanced paired data. The sequential design used would only result in confounding if the effect of time were somehow greater than drug effects. We contend that this is unlikely for the pain task data, as the fMRI response to repeated painful electrical stimulations recovers to the baseline magnitude of stimulus-induced signal change after a break of >4 min,^38,39 and our pain task blocks were separated by ∼15 min. Further, no systematic differences in resting connectivity measures would be expected in two sequential scans <20 min apart.

Increasing anxiety is a potential confound that could explain connectivity differences, and we did not quantify state of anxiety throughout the experiment. However, significant increases in anxiety levels are unlikely in this cohort of (mostly experienced) volunteer participants, especially given the absence of any reported anxiety or restlessness during or after the session. As a single agent study, we cannot directly dissociate the effects we observe that are specific to lidocaine from core changes due to analgesia by other means. Finally, we were unable to measure serial blood lidocaine concentrations, which would have confirmed that a state-state level had been reached. This concern is tempered by close adherence to a dosing schedule that predicts a constant effect-site concentration according to established pharmacokinetics.¹⁸ Any imprecisions in the pharmacokinetic model are likely to be overshadowed by pharmacodynamic variability between individuals.

Conclusions

In healthy adult volunteers, intravenous lidocaine administered at an effect site-concentration of 1.5 μg ml⁻¹ significantly affected fMRI measures of brain function, with minimal side effects. Lidocaine was associated with broad reductions in fMRI response to painful stimulation in regions commonly involved in acute pain processing, including the insula, cingulate gyrus, thalamus, and primary sensory cortex. Further, systemic lidocaine modulated whole-brain functional connectivity, predominantly decreasing long-range connectivity, with some predominance in temporal lobe structures. Small but significant decreases in pain ratings and cognitive task performance were also observed. This is consistent with lidocaine modifying the pain experience by affecting both primary sensory processing and higher-level processing of the noxious stimulus. This work lays the foundation for a better understanding of systems-level neuroscience changes that occur with lidocaine, working towards refining the clinical use of this important analgesic agent.

Authors’ contributions

Study design: KMV, JWI

Participant enrolment: MS, CNK

Data collection: KMV, MS, CNK

Data analysis: KMV, ACB, MS, SNR

Interpretation of results: KMV, ACB, JWI

Drafting manuscript: KMV, ACB, MS

Editing manuscript: KMV, ACB, MS, SNR, CNK, JWI

All authors meet the International Committee of Medical Journal Editors recommendations for authorship.

Declaration of interest

The authors declare that there are no conflicts of interest.

Funding

A research project grant (R35 GM146822) and Institutional Training grant (T32GM075770) from the National Institute of General Medical Sciences (NIGMS) of the US National Institutes of Health (NIH; Bethesda, MD, USA), and the Department of Anesthesiology and Perioperative Medicine at the University of Pittsburgh, School of Medicine (Pittsburgh, PA, USA).

Handling Editor: Hugh C Hemmings Jr

Appendix A. Supplementary data

The following is the Supplementary data to this article: