Mechanical pain sensitivity is associated with hippocampal structural integrity

Supplemental Digital Content is Available in the Text.

Individuals with structural hippocampal damage have lower mechanical pain sensitivity. The degree of structural damage is associated with the degree of mechanical hypoalgesia.

Abstract

Rodents and human studies indicate that the hippocampus, a brain region necessary for memory processing, responds to noxious stimuli. However, the hippocampus has yet to be considered a key brain region directly involved in the human pain experience. One approach to answer this question is to perform quantitative sensory testing on patients with hippocampal damage—ie, medial temporal lobe epilepsy. Some case studies and case series have performed such tests in a handful of patients with various types of epilepsy and have reported mixed results. Here, we aimed to determine whether mechanical pain sensitivity was altered in patients diagnosed with temporal lobe epilepsy. We first investigated whether mechanical pain sensitivity in patients with temporal lobe epilepsy differs from that of healthy individuals. Next, in patients with temporal lobe epilepsy, we evaluated whether the degree of pain sensitivity is associated with the degree of hippocampal integrity. Structural integrity was based on hippocampal volume, and functional integrity was based on verbal and visuospatial memory scores. Our findings show that patients with temporal lobe epilepsy have lower mechanical pain sensitivity than healthy individuals. Only left hippocampal volume was positively associated with mechanical pain sensitivity—the greater the hippocampal damage, the lower the sensitivity to mechanical pain. Hippocampal measures of functional integrity were not significantly associated with mechanical pain sensitivity, suggesting that the mechanisms of hippocampal pain processing may be different than its memory functions. Future studies are necessary to determine the mechanisms of pain processing in the hippocampus.

1. Introduction

Is the hippocampus—a structure known for memory and spatial navigation—critically involved in pain perception? There is limited discussion of hippocampal involvement in pain: <1% of preclinical and clinical studies have considered a role for the hippocampus in the pain experience.^3,54 However, previous animal studies have shown potential hippocampal contribution in nociception and pain sensation, including neural activity in response to noxious stimuli,^18,19,30 hippocampal involvement in antinociceptive mechanisms,^32,46,53 and reduced pain sensitivity (hypoalgesia) in nonhuman primates after hippocampal lesions.⁴⁰ In human neuroimaging studies, the hippocampus is activated in response to various acute noxious stimuli and in context-dependent experiments, such as pain expectancy and pain-related anxiety in healthy individuals,^3,12,38,39 suggesting a potential modulatory role for the hippocampus in pain. A study using intracranial neural recording in human patients with epilepsy showed activity in the hippocampus, as well as other more canonical pain-related regions, in response to a peripheral noxious laser stimulus.⁷ Although this study indicates hippocampal activation in response to noxious stimuli, evidence of direct hippocampal involvement in the human pain experience are necessary.

One approach to determine whether the hippocampus is directly involved in pain perception is to test pain sensitivity in individuals with hippocampal damage and dysfunction.^1,9,15,28,29 Specifically, these individuals have shown hypoalgesic responses to a noxious stimulus and painful conditions.^24,27 Temporal lobe epilepsy, a disorder with varying degrees of hippocampal abnormality, provides a unique opportunity to test the hypothesis that patients with temporal lobe epilepsy will show hypoalgesia that scales with the extent of hippocampal structural damage and dysfunction, demonstrating a link between the hippocampus and pain perception. To evaluate the extent of hippocampal abnormality, patients with temporal lobe epilepsy will undergo neuroimaging and neuropsychological evaluation of verbal and visuospatial memory as part of their standard clinical care, which evaluates structural damage and memory impairment, respectively.^49,60 Pain sensitivity, however, has yet to be evaluated using a standardized method in patients with temporal lobe epilepsy. Pain sensitivity can be assessed with quantitative sensory testing, a set of psychophysical evaluations that allow the detection of sensory impairment in the nociceptive system.⁴² Specifically, mechanical pain sensitivity (MPS) has yet to be performed on patients with temporal lobe epilepsy, which would evaluate pain intensity in response to mechanical pinprick stimuli.

Our first aim was to investigate whether patients with temporal lobe epilepsy have different pain sensitivity compared with healthy individuals. We hypothesized that patients with temporal lobe epilepsy would show hypoalgesic responses to MPS testing compared with healthy individuals. Our second aim was to investigate whether MPS is associated with measures of hippocampal integrity—ie, does the degree of damage affect the degree of hypoalgesia. This would provide evidence of direct hippocampal involvement in pain perception. We operationalize hippocampal structural integrity as hippocampal volume and functional integrity as memory function. Our study establishes the first direct evidence in humans associating the hippocampus to the pain experience.

2. Materials and methods

2.1. Participants

Our cross-sectional study was approved by the ethics committees at the University Health Network and the University of Toronto, and written informed consent was obtained from all participants. We recruited 19 patients with temporal lobe epilepsy and 19 age- and sex-matched healthy individuals. No participants used recreational drugs or pain medication (eg, acetaminophen or non-steroidal anti-inflammatory drugs) within 24 hours of testing. Localization and laterality of seizure onset was diagnosed based on clinical review of extended electroencephalographic and video monitoring data. Presence or absence of mesial temporal lobe sclerosis was determined by a neuroradiologist at the Toronto Western Hospital. Patients were seizure-free for at least 24 hours before testing and 17 were tested on the Epilepsy Monitoring Unit at which time they were being weaned off anti-seizure medications. Healthy individuals did not present with any pain, neurological, psychiatric, or other chronic conditions.

2.2. Mechanical pain sensitivity

All participants underwent quantitative sensory testing to MPS using PinPrick stimulators (MRC Systems GmbH, Heidelberg, Germany).⁴² Mechanical pain sensitivity was assessed on the volar forearm, bilaterally. First, we assessed the dominant arm, followed by the nondominant arm, with the order counterbalanced across participants. The PinPrick probes range from 8 to 512 mN increasing by log2, for a total of 7 stimulators. Mechanical pain sensitivity was evaluated across 5 runs with pseudorandom presentation of each PinPrick stimulator, as per the protocol defined by the German Research Network on Neuropathic Pain.^41,42 After each stimulus, the participant rated their pain intensity on a verbal numerical rating scale, with the following anchors: 0 = “no pain” and 100 = “the most painful imaginable.” For the first analysis, the arithmetic mean pain rating for each PinPrick stimulator was computed for each participant (10 measurements in total for each participant).

For the second analysis, the MPS variable entered in the linear mixed models reflected the arithmetic mean pain ratings in response to the 512 mN PinPrick for each limb. For 2 patients, the presence of an intravenous line prevented testing on that limb, such that only 5 measurements from the one forearm were available to compute MPS scores (See Supplementary Table 1, http://links.lww.com/PAIN/C29).

2.3. Measure of structural hippocampal damage in temporal lobe epilepsy

Patients underwent structural 3T T1-weighted magnetic resonance imaging (MRI) to obtain a measure of hippocampal damage based on volumetrics with FreeSurfer v7.2.²⁸ Healthy participants were not scanned because our second aim was to investigate the relationship between hippocampal integrity and pain sensitivity. Magnetic resonance imaging acquisition parameters are presented in Supplementary Table 2, http://links.lww.com/PAIN/C29. Briefly, all T1-weighted scans underwent “recon-all” pipeline, which preprocesses, normalizes, segments, and labels brain regions and computes the morphometric values for each brain structure.²¹ This process was followed by a hippocampal and amygdala segmentation algorithm module implemented in FreeSurfer.²⁸ The probabilistic atlas for this module was constructed using a Bayesian algorithm with high-resolution ex vivo (0.1 mm) and in vivo (1 mm) MRI data to generate an automated segmentation of the hippocampus.^28,44 All segmentations ran without error. Quality control was performed using the Enigma protocols for the hippocampal subfield module (www.enigma.ini.usc.edu).⁴⁴ First, an outlier detection algorithm was performed across the entire group. This algorithm identifies whether patients' right and left hippocampal volumes and estimated total intracranial volumes are larger or smaller than 2 SDs outside the cohort mean. Volumetric outliers were not detected. Each patient's left and right hippocampal segmentations were also overlayed onto the bias-corrected T1-weighted scan and were inspected visually by L.J.A. and M.P.M. for good positioning and segmentation coverage. One patient scan was excluded after visual inspection because of low image quality due to technical error. An example of a hippocampal segmentation in a patient with left temporal lobe epilepsy and left mesial temporal lobe sclerosis is provided in Supplementary Fig. 1, http://links.lww.com/PAIN/C29. For each patient, left and right hippocampal volumes were each normalized with the estimated total intracranial volume using the normalized ratio method (hippocampus volume/estimated total intracranial volume).^4,52,59 Smaller normalized values represent greater hippocampal atrophy.⁵ Given the role of the amygdala in pain,^36,63 and the dense connectivity between the amygdala and the hippocampus,⁶⁴ we performed a control analysis with the left and right amygdalar volumes, normalized with the estimated total intracranial volume.

2.4. Measures of hippocampal dysfunction in temporal lobe epilepsy

Patients also underwent neuropsychological testing as part of their clinical evaluation: verbal and visuospatial memory were assessed. Verbal and visuospatial memory indices were computed according to a principal components analysis, as performed in our previous work.⁴⁹ The resulting verbal memory index was more weighted by the Warrington's Word Recognition Test and the Rey Auditory Verbal Learning Test scores. The Warrington's Word Recognition Test involved the presentation of 50 words followed by a 2-alternative forced choice recognition test.⁵⁶ The score was the total number of correct recognition responses. The Rey Auditory Verbal Learning Test consists of presenting 15 words for free recall over 5 learning trials, followed by a distractor list for 1 trial, and then immediate and 20-minute delayed recall trials. The scores contributing to this index are the total correct responses over 5 learning trials and percent retained after the 20-minute delay.⁵⁰ The visuospatial memory index was more weighted by the Warrington Face Recognition test, the Rey Visual Design Learning Test, and the Spatial Conditional Associative Learning task. The Warrington Face Recognition test involved the presentation of 50 faces followed by a 2-alternative forced-choice recognition test.⁵⁶ The total score was the number of correct recognition responses. The Rey Visual Design Learning Test consists of presenting 15 drawings for 5 free recall trials, and the score was the total correct responses over the 5 learning trials.⁴⁷ The Spatial Conditional Associative Learning task involved learning a set of 4 spatial associations.⁵¹ The score was the total number of trials needed to reach a criterion of 12 consecutive correct responses. The verbal and performance intellectual quotient from the Wechsler Abbreviated Scale of Intelligence⁵⁷ were also evaluated in the principal components analysis but do not load heavily onto the verbal memory index or the visuospatial memory index. These memory index scores can be interpreted as pseudo-z scores, with higher numbers representing better performance. Each patient had one verbal memory and one visuospatial memory score. Supplementary Fig. 2, http://links.lww.com/PAIN/C29, provides the equations used to compute the scores for each memory index.⁴⁹

2.5. Analysis

In the first analysis, to determine whether patients with temporal lobe epilepsy had lower MPS than healthy individuals, we evaluated MPS differences between patients with temporal lobe epilepsy and healthy individuals. In GraphPad Prism v.9.5.1 (GraphPad Software, Boston, MA), we performed a 2-way repeated-measures ANOVA to compute differences between groups across different levels of PinPrick weight and corrected for multiple comparisons using the Bonferroni method, with significance set at a corrected P < 0.05.

To address our second aim, we evaluated whether the extent of mechanical hypoalgesia is related to the extent of hippocampal integrity in patients. These analyses were performed using the lme4 package in RStudio v.4.2.2.^8,43 We performed 3 linear mixed models to determine whether there was a significant relationship between pain intensity ratings elicited by the 512 mN PinPrick (which served as a measure of mechanical pain) and each measure of hippocampal integrity (normalized volume, verbal memory, visuospatial memory). Specifically, we performed 3 separate linear mixed models with MPS pain intensity ratings as the dependent variable and modelled a random-intercept for each subject. In these models, we controlled for MPS laterality (whether MPS ratings are rated from the left and right forearm stimulation), seizure onset laterality, and number of anti-seizure medications taken by each patient. All predictors were centered and standardized before model estimation. We included anti-seizure medications, because of their potential analgesic effects that would lower pain sensitivity,⁴⁵ and seizure laterality given the diagnostic heterogeneity in our patient sample. Significance was set at P < 0.05.

In the first model, the left normalized hippocampal volume and its interaction with MPS laterality, the right normalized hippocampal volume and its interaction with MPS laterality were entered as independent variables, along with seizure laterality, the number of medications, and MPS laterality.

$$MPS\hspace{0.17em}\sim Left\hspace{0.17em}Hippocampal\hspace{0.17em}Volume+Left\hspace{0.17em}Hippocampal\hspace{0.17em}Volume\times MPS\hspace{0.17em}Laterality+Right\hspace{0.17em}Hippocampal\hspace{0.17em}Volume+Right\hspace{0.17em}Hippocampal\hspace{0.17em}Volume\times MPS\hspace{0.17em}Laterality+Seizure\hspace{0.17em}Laterality+Number\hspace{0.17em}of\hspace{0.17em}Medications+MPS\hspace{0.17em}Laterality+\left(1|Subject\right)$$ (1)

Equation 1: Model testing the relationship between MPS and hippocampal volume.

The second model included verbal memory scores and their interaction with MPS laterality as independent variables along with seizure laterality, the number of medications, and MPS laterality.

$$MPS\hspace{0.17em}\sim Verbal\hspace{0.17em}Memory+Verbal\hspace{0.17em}Memory\times MPS\hspace{0.17em}Laterality+Seizure\hspace{0.17em}Laterality+Number\hspace{0.17em}of\hspace{0.17em}Medications+MPS\hspace{0.17em}Laterality+\left(1|Subject\right)$$ (2)

Equation 2: Model testing the relationship between MPS and verbal memory.

The third model included visuospatial memory scores and their interaction with MPS laterality as independent variables along with seizure laterality, the number of medications, and MPS laterality.

$$MPS\hspace{0.17em}\sim Visuospatial\hspace{0.17em}Memory+Visuospatial\hspace{0.17em}Memory\times MPS\hspace{0.17em}Laterality+Seizure\hspace{0.17em}Laterality+Number\hspace{0.17em}of\hspace{0.17em}Medications+MPS\hspace{0.17em}Laterality+\left(1│Subject\right)$$ (3)

Equation 3: Model testing the relationship between MPS and visuospatial memory.

As a control analysis, we performed a fourth model that included the left normalized amygdala volume and its interaction with MPS laterality, the right normalized amygdala volume and its interaction with MPS laterality as independent variables, with seizure laterality, medication, and MPS laterality.

$$MPS\hspace{0.17em}\sim Left\hspace{0.17em}Amygdala\hspace{0.17em}Volume+Left\hspace{0.17em}Amygdala\hspace{0.17em}Volume\times MPS\hspace{0.17em}Laterality+Right\hspace{0.17em}Amygdala\hspace{0.17em}Volume+Right\hspace{0.17em}Amygdala\hspace{0.17em}Volume\times MPS\hspace{0.17em}Laterality+Seizure\hspace{0.17em}Laterality+Number\hspace{0.17em}of\hspace{0.17em}Medications+MPS\hspace{0.17em}Laterality+\left(1|Subject\right)$$ (4)

Equation 4: Model testing the relationship between MPS and amygdala volume.

3. Results

Participant demographics along with measures of pain intensity, hippocampal raw volumes, normalized hippocampal volumes, and memory index scores are presented in Table 1. Further detail on patient clinical characteristics and their FreeSurfer volumetric outputs for the hippocampus, amygdala, and estimated total intracranial volumes (mm³) is provided in Supplementary Table 1, http://links.lww.com/PAIN/C29. Patients had lower MPS pain ratings compared with healthy individuals, supported by a significant interaction between PinPrick weight and group (F_{6, 216} = 3.80, P = 0.0013); see Figure 1. Post hoc tests indicated that the group differences were observed for 256 mN (adjusted-P = 0.044) and 512 mN (adjusted-P = 0.014).

Table 1.

Participants included in the study.

	Age (mean ± SD in years)	Sex (F:M)	MPS at 512 mN* (/100)	Measures of hippocampal integrity (mean ± SD)
				L hippocampus		R hippocampus		VSM	VM
				Volume (mm3)	Normalized	Volume (mm3)	Normalized
TLE (n = 19)†	34.8 ± 10.4	7:12	6.72 ± 9.88	3434.45 ± 469.88	0.0021 ± 0.00042	3562.57 ± 464.07	0.0022 ± 0.00037	−0.24 ± 0.83	−0.15 ± 1.04
Healthy (n = 19)	33.8 ± 11.0	7:12	16.95 ± 21.44	—	—	—	—	—	—

^*MPS rating for n = 18 patients (excluding n = 1 because of MRI data) is 7.04 ± 10.07 (mean ± SD).

^†One patient was excluded from the second analysis because of the quality of the MRI data.

F, female; L, left; M, male; MPS, mechanical pain sensitivity; R, right; TLE, temporal lobe epilepsy; VM, verbal memory; VSM, visuospatial memory.

Figure 1.

Patients with temporal lobe epilepsy are hypoalgesic to mechanical stimuli in the noxious range. Patients have significantly less pain sensitivity compared with healthy individuals (F_{6, 216} = 3.80, P = 0.0013), especially for the heaviest PinPrick weights (256 mN and 512 mN) (both adjusted-P < 0.05). Data points and error bars represent the mean and SD, respectively. TLE, temporal lobe epilepsy.

In patients, we observed a significant positive effect of left normalized hippocampal volume on MPS pain rating (t (14.4) = 3.02, P = 0.009), indicating that a smaller left normalized hippocampus is associated with lower MPS pain ratings, see Figure 2 and Table 2. We did not observe this relationship with the right normalized hippocampal volume (t (17.4) = −0.5, P = 0.65) nor any interaction between normalized hippocampal volumes and MPS laterality (all t < 0.2, P > 0.8). We additionally did not observe a significant effect of MPS laterality (t (13.6) = −0.54, P = 0.6), but we did see a trending effect of medication (t (11.9) = −2.02, P = 0.07) in which a higher number of medications taken was associated with lower MPS pain ratings. Finally, we also observed a marginal effect of seizure laterality with right hemisphere seizure onset being associated with lower MPS pain ratings compared with left hemisphere onset (t (11.9) = 2.02, P = 0.07). Models testing functional hippocampal integrity (ie, verbal memory and visuospatial memory) presented in Table 3 were not significant; verbal memory: t (14.5) = 0.21, P = 0.84; and visuospatial memory: t (14.6) = 0.87, P = 0.4. Models testing the normalized amygdala were also not significant (left normalized amygdala: (t (13.7) = 1.53, P = 0.15; right normalized amygdala: t (14.2) = −0.61, P = 0.55; See Supplementary Table 3, http://links.lww.com/PAIN/C29).

Figure 2.

Structure, but not function, is significantly associated with pain sensitivity. (A) The left normalized hippocampus is positively associated with pain ratings to the 512 mN PinPrick; (B) visuospatial memory and (C) verbal memory are not significantly associated with pain ratings from the 512 mN PinPrick. Values represented are standardized z-scores of the residuals from their respective linear mixed models, such that a value of 0 represent the average and a positive or negative value represent above and below average, respectively. The 95% confidence intervals are shown in grey. HC, hippocampus; MPS, mechanical pain sensitivity.

Table 2.

Model test statistics for the relationship between the hippocampus volume and pain sensitivity.

Term name	t	df	P
Left normalized hippocampus	3.02	14.4	0.009
Right normalized hippocampus	−0.5	17.4	0.65
MPS laterality	−0.54	13.6	0.6
Medication	−2.02	11.9	0.07
Seizure laterality (bilateral)	−0.25	12.6	0.81
Seizure laterality (unilateral)	2.02	11.9	0.07
Left normalized hippocampus × MPS laterality	−0.043	13.5	0.97
Right normalized hippocampus × MPS laterality	0.24	13.7	0.81

A linear mixed model of the relationship between the hippocampus volume and mechanical pain sensitivity rating, with factors such as the number of anti-seizure medications, mechanical pain sensitivity laterality (coded for the left and right forearm), and seizure laterality (bilateral seizure onset for bilateral vs right coded as [0,−1] and unilateral seizure onset for left vs right coded as [1,−1]).

Significance was set at P < 0.05.

MPS, mechanical pain sensitivity.

Table 3.

Model test statistics for the relationship between hippocampal functional integrity and pain sensitivity.

Term name	t	df	P
Visuospatial memory model
VSM	0.87	14.6	0.4
MPS laterality	−0.39	14.3	0.7
Medication	−0.87	12.9	0.41
Seizure laterality (bilateral)	0.29	13	0.78
Seizure laterality (unilateral)	0.94	12.9	0.37
VSM × MPS laterality	1.46	14.5	0.17
Verbal memory model
VM	0.21	14.5	0.84
MPS laterality	−0.54	14.2	0.6
Medication	−0.68	12.9	0.51
Seizure laterality (bilateral)	0.02	12.8	0.99
Seizure laterality (unilateral)	0.91	12.9	0.38
VM × MPS laterality	1.4	14.2	0.2

Linear mixed models of the relationship between hippocampal functional integrity and mechanical pain sensitivity rating, with factors such as the number of anti-seizure medications, mechanical pain sensitivity laterality (coded for the right and left forearm) and seizure laterality (bilateral seizure onset for bilateral vs right coded as [0,−1], and unilateral seizure onset for left vs right coded as [1,−1]).

MPS, mechanical pain sensitivity; VM, verbal memory; VSM, visuospatial memory.

4. Discussion

Our study establishes the first direct association between the hippocampus and MPS in humans. Patients with temporal lobe epilepsy showed a rightward shift of the stimulus–response curve in MPS, indicating hypoalgesia, compared with healthy individuals. Furthermore, patients with greater extent of hippocampal damage, specifically the left hippocampus, had lower MPS. Our findings show, for the first time, that pain perception is associated with the integrity of the hippocampus in humans.

In line with our findings, previous case studies identified hypoalgesia in patients with hippocampal damage.^24,27 Because these studies evaluated patients after neurosurgery, ie, resection of the hippocampus and neighbouring structures including the amygdala, it is impossible to disentangle whether hypoalgesia was related to hippocampal resection—especially given the role of the amygdala in pain perception and modulation.⁶³ Here, we provide novel evidence by quantifying the extent of structural damage in the hippocampus and identifying its relationship with pain sensitivity. How the hippocampus is involved in the pain experience and in what capacity remains to be fully characterized. An acute painful stimulus will engage ascending nociceptive circuits, where nociceptive input is transmitted and processed in several subcortical and cortical regions, as well as descending modulatory circuits.^16,61,62 Reduced pain sensitivity in patients with temporal lobe epilepsy indicates that the hippocampus is integral to the pain experience and perhaps even necessary for maintaining network integrity with other brain regions involved in pain processing, but this has yet to be determined. The left hippocampal volume significantly contributed to pain sensitivity irrespective of whether testing was performed on the right or left forearm in patients. It is feasible that in our patients with unilateral damage, the unaffected hippocampus supported their ability to feel pain, but clearly the intensity of the pain experience was reduced in proportion to the degree of damage in the hippocampus. It remains unknown, however, how the hippocampus processes pain sensation, and from which brain regions it receives nociceptive information.

We did not find a relationship between our measures of functional integrity (eg, verbal and visuospatial memory scores) and pain. A few previous studies have investigated the relationship between memory and pain. In healthy individuals, subsequent recall was reduced for images paired with thermal pain,²³ which was interpreted as a pain-related disruption during memory encoding.²² The relationship between autobiographical memory and pain was explored in a longitudinal study of patients undergoing major surgery.⁵⁵ Those who took longer to recall pain-specific autobiographical memory presurgery were more likely to develop chronic postsurgical pain, suggesting specific memory retrieval might inoculate one to subsequent pain experiences. However, it remains unclear whether and how the hippocampus itself is driving the association between acute pain sensitivity and memory. The lack of an association between our neuropsychological memory variables and pain sensitivity could reflect their reliance on the integrity of brain networks beyond the hippocampus.⁶ Although we have documented associations between these index scores derived from standard clinical assessments and hippocampal volume and functional activation in past work,⁵ we do not assume they are influenced exclusively by hippocampal integrity. It remains possible that measures more specific to known hippocampal functions such as pattern separation⁶⁵ might provide a stronger link between hippocampally based memory functions and pain, or that alternate ways of accessing pain-relevant memory contents are required to identify these links. However, the hippocampus could have a mechanism specific to pain sensitivity that is unrelated to memory mechanisms altogether.

How the hippocampus computes incoming information about an injury and how it interacts with other brain regions are important inquiries towards understanding its contribution to pain. The experience of pain involves an ensemble of brain regions and networks contributing to the sensory-discriminative, motivational-affective, and cognitive-evaluative domains.³³ One study reported that patients with temporal lobe epilepsy had reduced leg flexion nociceptive reflex (a RIII reflex) compared with healthy individuals, which was not reversed by naloxone.²⁶ This suggests that this underlying mechanism is not opioidergic. In another study, activity in the hippocampus and central nucleus of the amygdala were recorded when patients received noxious laser stimulation, indicating that these structures receive nociceptive input or input from other brain areas—we do know that there is activity in response to a noxious stimulus.³¹ Nociceptive information reached the hippocampus 150 to 180 milliseconds after the noxious stimulus delivery compared with 120 milliseconds for regions thought to process low-level nociceptive information, such as the posterior insula.⁷ The hippocampus was further investigated in human studies involving the affective and cognitive domains of pain, given its reciprocal projections to the amygdala and specific contribution to emotional processing.³⁵ In healthy individuals, increased nocebo effect (negative expectancy to remifentanil, an opioid agonist) was associated with left hippocampal activation,¹¹ and its increased functional connectivity with brain regions involved in the sensory-discriminative dimension of pain (eg, primary somatosensory cortex and posterior insula), and affective and pain modulatory regions (eg, amygdala and periaqueductal gray).¹² Specifically, increased pain intensity in the nocebo condition was related to weaker functional connectivity between the right amygdala and left hippocampus. These results support previous studies investigating hippocampal engagement in context-dependent pain experiments involving many cognitive processes, such as pain expectancy and pain amplification.^38,39 Importantly, studies including both healthy and damaged hippocampi could further unravel the extent of its contribution to pain sensitivity directly.

Hippocampal lesion studies could be further pursued to characterize functional specificity within the hippocampus and its network. Previous evidence in chronic pain models in rodents have shown such differentiation: the dorsal (posterior hippocampus in humans) and not the ventral hippocampus (anterior hippocampus in humans) contributed to pain reduction in rodents.⁵⁸ Importantly, localizing lesions within the hippocampus could inform whether and how the anterior or posterior hippocampus may affect pain experience. One approach could be to use diffusion weighted imaging to evaluate structural integrity.¹⁴ There is also an outstanding gap in the literature regarding the characterization of hippocampal projections with the amygdala and other limbic regions, and how these regions interact to shape the pain experience. Differences in pain behaviour in rodents with amygdalar or hippocampal lesions would provide much needed mechanistic information on their roles in pain processing.

The hippocampus is an important brain region to investigate in the context of pain because of previous evidence of its involvement in acute pain and its role in the development of chronic pain. In our previous meta-analysis, the right hippocampus was activated in response to experimental pain and showed less activity and reduced functional network in patients with chronic pain compared with healthy individuals.³ Previous studies also found hippocampal involvement in the development of chronic pain^2,34,54 and normalization of hippocampal volume after pain relief after neurosurgery in patients with trigeminal neuralgia.³⁷ Our findings complement these studies in a unique patient population which presents with specific hippocampal abnormalities and proposes a novel model of hippocampal abnormality in the investigation of acute pain processing in humans.

4.1. Limitations

Lesion studies hold several limitations because of the variability in brain lesion location and small sample sizes because of patient scarcity. The strength of our study rests on the specificity of patient diagnosis, such that lesions directly affect the hippocampus, and patient recruitment was based in one of the largest epilepsy monitoring units in Canada. We acknowledge that our sample is skewed by individuals with left temporal lobe epilepsy, and thus, this tempers any conclusions of left hippocampal exclusivity in the relationship with hypoalgesia. Given our small sample size and the extensive variability in temporal lobe epilepsy diagnoses (seizure laterality on the right, left or bilaterally) and presence or not of damage (mesial temporal sclerosis on the right, left or bilaterally), further studies should investigate direct laterality differences with a robust sample size. Previous case studies in patients with brain lesions involved in pain processing, eg, insula, amygdala, mid-cingulate cortex, and in those after resections of white matter tracts, eg, cingulum, have also shown altered pain responses to noxious stimuli with relatively low sample sizes (n < 10).^{10,17,20,25,48} In addition, our results should be interpreted with the following limitation: the association between hippocampal integrity and pain sensitivity may not infer causality. However, we do provide direct evidence for the relationship between pain sensitivity and hippocampal structural integrity. Overall, these are compelling evidence that pain perception is a highly conserved brain phenomenon, as shown here and others previously.¹³ Finally, we did not evaluate sex differences given our small sample size. Importantly, large sample size studies will determine whether there are sex differences in hippocampal-pain relationships.

In conclusion, we propose that the hippocampus plays an important role in the experience of pain—damage to the left hippocampus blunts pain responses and the extent of this deficit is related to the extent of hippocampal damage. These findings are foundational evidence for future mechanistic studies unravelling the role of the hippocampus in pain.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Appendix A. Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C29.