Sex Differences in Brain Regions Modulating Pain Among Older Adults: A Cross-Sectional Resting State Functional Connectivity Study

Abstract

Objective

A long-standing hypothesis is that when compared with males, females may be at increased risk of experiencing greater pain sensitivity and unpleasantness. The purpose of this study was to examine sex differences in pain psychophysics and resting state functional connectivity (RSFC) in core pain regions in an age- and sex-matched sample of healthy older adults.

Design

Between groups, cross-sectional.

Setting

Vanderbilt University and Medical Center.

Subjects

The sample in the analyses reported here consisted of 19 cognitively intact males matched with 19 cognitively intact females of similar ages (median ages: females = 70 years, males = 68 years).

Methods

Psychophysical assessment of experimental thermal pain and RSFC.

Results

There were no significant differences in perceptual thresholds or unpleasantness ratings in response to thermal stimuli. Older males showed greater RSFC between the affective and sensory networks and between affective and descending modulatory networks. Conversely, older females showed greater RSFC between the descending modulatory network and both sensory and affective networks. The strongest evidence for sex differences emerged in the associations of thermal pain with RSFC between the anterior cingulate cortex (ACC) and amygdala and between the ACC and periaqueductal gray matter in older females relative to older males.

Conclusions

We found no differences in pain sensitivity or pain affect between older males and older females. Additionally, we found that older females exhibited a greater association between thermal pain sensitivity and RSFC signal between regions typically associated with pain affect and the descending modulatory system. One interpretation of these findings is that older females may better engage the descending pain modulatory system. This better engagement possibly translates into older females having similar perceptual thresholds for temperature sensitivity and unpleasantness associated with mild and moderate pain. These findings contrast with studies demonstrating that younger females find thermal pain more sensitive and more unpleasant.

Introduction

Untreated acute and chronic pain are typically the most common reasons people seek medical care [1], and poorly treated pain is associated with depression, loss of sleep, functional decline, delayed healing, and even hastened death [2]. The risk of developing illnesses that may lead to chronic pain increases with age, and suffering from pain can profoundly impact the quality of life in older adults [3]. One factor that may influence the response to and management of pain in older adults is sex differences. Sex differences in pain processing and management have been extensively documented in younger adults and to some degree in older adults as well [4]. Relative to males, females have a greater incidence of chronic and postprocedural pain while demonstrating increased sensitivity and unpleasantness to various types of experimental pain (reviewed in [5]). Moreover, among middle- and older-aged adults with osteoarthritis, relative to males, females were more pain sensitive [6]. Findings from a recent meta-analysis demonstrate that when compared with males, females appear to show a greater analgesic response to morphine [7]. Neuroplastic alterations in response to pain are known to occur throughout the lifetime. Females have greater exposure to pain episodes over their lives [8], and it is plausible that repeated exposure to pain might promote neuroplastic changes that differentially alter the sensory and affective pain networks in the brains of older females. These changes might, in turn, alter pain perception or reporting and impact treatment. The biologic reasons for these sex differences remain poorly understood [9], especially in older adults because most studies examining sex differences in pain are of people younger than 65 years of age (reviewed in [5,10]). However, lack of understanding of potential sex differences in pain processing in older adults may result in inadequate and poorly targeted treatments for specific individuals.

Central pain processing is complicated and involves a multitude of brain regions working in concert [11]. Pain is an unpleasant sensory and emotional experience [12], likely mediated partially through sensory and affective pain networks [13]. Core sensory network regions include primary (S1) and secondary (S2) somatosensory cortices [14], as well as the posterior insular cortex (pINS) [15], while affective regions responsible for encoding unpleasantness include subgenual anterior cingulate cortex (sgACC) [16], dorsolateral prefrontal cortex (dlPFC) [17], and anterior insular cortex (aINS) [18].

In addition to the brain regions comprising the sensory and affective networks, pain processing is influenced by other modulatory systems. The periaqueductal gray matter (PAG) is a key structure in the brain responsible for pain modulation [19] through its descending cortical connections with the rostral ventromedial medulla (RVM) [20]. In addition to its connections to the RVM, the PAG is one component of a larger network comprising the descending modulatory system that also includes the hypothalamus (HYPO), amygdala (AMY), insula (INS), and ACC (reviewed in [11]). The PAG, ACC, and RVM form a core, functionally connected network [21]. Though previous findings are somewhat mixed, multiple studies suggest the importance of the PAG and its functional connections with sensory and affective pain regions as contributors to sex differences in pain [22,23].

One method for demonstrating differences in interactions between pain network regions in humans is resting state functional magnetic resonance imaging (fMRI). Resting state fMRI is used to describe temporal correlations of activations across brain regions, which may reflect both functional and structural connectivity and brain function [24]. Recently, resting state functional connectivity (RSFC) analysis was used to demonstrate sex differences in pain modulation in younger adults age 19 to 36 years [25]. For example, increased RSFC between sgACC and PAG in females may be associated with increased pain habituation when compared with males [23] or greater RSFC between PAG. AMY, caudate, and putamen may result in an increased ability to modulate pain in males [26].

In younger adults, RSFC studies have revealed central mechanistic sex differences in painful conditions such as irritable bowel syndrome [27], migraine [28], and chronic abdominal pain [29]. Interestingly, longitudinal RSFC has been used to describe central alterations associated with chronic pain. Among people with newly diagnosed subacute back pain (SABP), RSFC was used to differentiate individuals who transitioned from SABP to chronic low back pain from those who did not [30], underscoring the potential clinical and translational relevance of using RSFC in the diagnosis and management of chronic pain.

At present, the bulk of the literature on RSFC and pain thus far is derived from younger cohorts, with limited studies examining RSFC and pain in older populations [31,32]. As such, there is a critical need to use RSFC to examine sex differences in pain processing in older adults. Thus, we designed our study to measure the psychophysical response to experimental thermal pain concerning a perceptual threshold for pain detection (sensory) and the degree of self-reported unpleasantness associated with the pain (a measure of affective response) in order to measure RSFC between core pain networks in an age-matched sample of cognitively intact, healthy older adults of equal sex distribution. The current sample was drawn from a larger ongoing study examining brain activation and pain reports in response to experimental thermal pain in older adults. In a pilot sample, we recently reported sex differences in the psychophysical and neurophysiological (fMRI task) response to pain in a sample of 24 older adults (50% female) [33]. Results demonstrated that older females found pain to be less unpleasant and exhibited lower pain detection temperatures than men (increased pain sensitivity) for both mild and moderate pain, suggesting a greater ability to detect and identify painful stimuli than their male counterparts. Contrary to our hypotheses, males exhibited greater activation to moderate pain in lateral (sensory/discriminatory) pain pathway regions than did females.

The current study extends this work by examining the association of psychophysics with RSFC in a larger age-matched sample of 38 older adults (50% female) who completed both psychophysics and resting state procedures in the parent study at the time of this analysis.

Our first hypothesis (psychophysical) was that compared with older males, older females would report increased pain intensity (greater pain for a given temperature) and increased unpleasantness to the pain of equal intensity. Our second hypothesis (neurophysiological) was that RSFC between sensory, affective, and modulatory pain regions would differ by sex. We selected the sgACC as one component of the affective network in particular because of its cortical connections with the descending modulatory network [23]. Because experimental and clinical pain studies typically find that women are more sensitive to pain [5], we hypothesized that when compared with older males, older females would demonstrate increased RSFC between the dlPFC, sgACC, aINS (affective network) and S1, S2, midinsula (mINS), pINS (sensory network). We further posited that when compared with older females, older males would demonstrate greater RSFC between the S1, S2, mINS, pINS (sensory network), and the AMY, HYPO, and PAG (descending modulatory network). Our third hypothesis was that RSFC would correlate with psychophysical responses.

Methods

Study Population

The Vanderbilt University Institutional Review Board approved the study, and all participants provided written informed consent. Participants were reimbursed $100.00 for their time. Of the 24 subjects reported in the original pilot study [33], nine participants are not included in this study because they had missing or incomplete resting state scans, which occurred after the functional task portion of the paradigm. Data from the remaining 23 patients in the current sample were acquired after the publication of the pilot study. The current sample was a secondary analysis from our originally designed study, which was powered to detect differences in people with Alzheimer’s disease (AD) relative to controls. However, a recent meta-analyses of neuroimaging studies examining sex differences in bilateral language representation found that 17 of the 24 studies included in the analyses were equal to or less than the sample size used in the current analyses [34]. In the current analyses, we matched 19 cognitively intact males with 19 cognitively intact females of similar ages (median ages: females = 70 years, males = 68 years). The recruitment procedures are described in two papers, one explaining sex differences in pain reports and fRMI pain task responses in cognitively intact controls [33] and the second reporting the psychophysical response to thermal pain in participants with and without Alzheimer’s dementia [35]. In brief, participants were recruited between 2012 and 2015 and were excluded for presence of chronic pain; cognitive impairment (Mini-Mental State Exam [MMSE] < 28) [36]; daily use of analgesic medication; or history of stroke, cancer, peripheral neuropathy, Raynaud’s Disease, unstable conditions (e.g., severe restrictive or obstructive lung disease), insulin-dependent diabetes, current substance use disorders, bipolar disorder, major depressive disorder, or schizophrenia.

After obtaining informed consent and subject assent, further demographic and health history information was obtained (e.g., age, race, height, weight, medical history), as well as metal screening for MRI safety clearance. Because the participants in this study consisted of very healthy older adults, most subjects were not taking medications. However, four general classes of medications were identified in the participants: 1) vitamins, 2) supplements, 3) blood pressure, and 4) cholesterol medications. There were no differences by group. Lastly, the Hollingshead four-factor index of socio-economic status (SES) [36] was administered. On the day of the MRI procedures, each subject completed the Brief Pain Inventory (BPI) [37], Geriatric Depression Scale (GDS) [38], and the State-Trait Anxiety Inventory (STAI) [39]. All participants were instructed to avoid drinking caffeine for four hours before scanning and not to use any pain medication (opioid or nonopioid) for at least 24 hours before data collection.

Psychophysical Assessments

On the day of the MRI, in an experimental room adjacent to the scanner, participants completed a thermal pain psychophysics evaluation (∼30 minutes) using the Method of Limits Program available in the Medoc CHEPS Pathway Pain and Sensory Evaluation System [40]. For the Method of Limits, the subject is exposed to a stimulus of changing/increasing intensity and asked to indicate the first onset of sensation. For this experiment, the stimulus increased by 1°C/second. These procedures are described in detail in a recent publication from the parent study examining thermal pain thresholds in people with Alzheimer’s dementia [35]. Briefly, participants were instructed to identify temperatures they perceived as “warmth,” “mild pain,” and “moderate pain.” Then, participants were shown a 0 to 20 pain unpleasantness scale used in prior work in older adults [41,42], then asked to rate the unpleasantness associated with each temperature percept. After three trials targeting each percept, the average temperature for warmth detection threshold, mild pain detection threshold, and moderate pain detection threshold, as well as the average unpleasantness associated with each, was recorded.

Brain Imaging Acquisition

Brain images were acquired on a Philips 3T Achieva MRI scanner (Philips Healthcare Inc., Best, the Netherlands). A standard whole-brain 3D anatomical T1-weighted turbo field echo (TFE with SENSE coil) scan was acquired. During each 300-second duration resting state functional run, 57 echo planar images (EPI) were acquired: 150 dynamics, 4.00 mm slice thickness with 0.40 mm gap, two-second time to repeat (TR), 35 ms echo time (TE), 79° flip angle, field of view (FOV) = 240 × 240 mm, reconstruction matrix = 128 × 128, acquisition resolution = 80 × 79. Before the acquisition of the resting state sequence, participants received randomly delivered thermal stimuli of “warmth,” “mild pain,” and “moderate pain” during four 16-minute fMRI blood oxygenation level–dependent (BOLD) task runs (results reported in [33]). For the five-minute resting state sequence in the current study, subjects were instructed to rest quietly with eyes open but to avoid falling asleep. Research assistants confirmed that each subject was awake at the end of the scan. A high-resolution T1-weighted structural scan was also acquired. Because our pilot data demonstrated that older adults can become fatigued or uncomfortable lying in the scanner, we prioritized the collection of the fMRI task data prior to the collection of the resting state sequence.

Definition of Regions of Interest

The literature on pain systems describes many regions responsible for processing the experience of pain. The regions of interest (ROIs) in the current analyses were chosen based on our own comprehensive review [43] and several other review articles [11,14]. While we could have increased the number of ROIs, we selected three to help limit the number of comparisons and to aid in the interpretation of potentially complicated relationships between the regions selected. The PAG was defined as a single sphere with a diameter of 6 mm centered on Montreal Neurological Institute (MNI) coordinates x = 6, y = –30, z = 14 [44]. We subdivided the INS into left aINS = 4 mm sphere centered on MNI coordinates x = –32, y = 22, z = 0; right aINS = 4 mm sphere centered on MNI coordinates x = 38, y = 16, z = 0; left mINS= 4 mm sphere centered on MNI coordinates x = –36, y = 12, z = 0; right mINS = 4 mm sphere centered on MNI coordinates x = 38, y = 12, z = 2; left pINS= 4 mm sphere centered on MNI coordinates x = –38, y = –16, z = 8; and right pINS = 4 mm sphere centered on MNI coordinates x = 38, y = –16, z = 8 [44]. The sgACC regions were adopted from Wang and colleagues [23] exploring the functional connectivity between the sgACC and descending modulatory regions, defined as six bilateral regions drawn as 3 mm spheres centered on MNI coordinates: 1) x = –5, y = 25, z = 10; x = 5, y = 25, z = 10; 2) x = –5, y = 34, z = 9; x = 5, y = 34, z = 9; 3) x = –6, y = 33, z = 9; x = 6, y = 33, z = 9; 4) x = –5, y = 34, z = 4; x = 5, y = 34, z = 4; 5) x = –6, y = 27, z = 10; x = 6, y = 27, z = 10; 6) x = –6, y = 30, z = 9; x = 6, y = 30, z = 9 (23). Next, we used Wake Forest University PickAtlas [45,46] to generate anatomical ROIs for S1 (BA 1, 2, 3), S2 (BA 40, 44), dlPFC (BA 9, 46), HYPO, and AMY. See Supplementary Table 1 for sources and atlases used to generate ROIs.

Preprocessing and Resting State Functional Connectivity Analysis

Preprocessing included motion correction, slice timing correction, band-pass filtering of 0.01 to 0.1 Hz, coregistration to structural images, spatial normalization to MNI space, and spatial smoothing with an 8 mm Gaussian kernel. Each subject’s average white matter, cerebrospinal fluid, and average motion parameters (in mm) were included in the first level model as variables of no interest. The CONN-fMRI Functional Connectivity Toolbox (v.15.g; www.nitrc.org/projects/conn/), as cited in Whitfield-Gabrieli and Nieto-Castanon [47] and CompCor [48], was used to perform these analyses.

We used an a priori ROI-to-ROI approach with the goal of determining sex differences in the RSFC within and between three networks: 1) the descending modulatory network (AMY, HYPO, and PAG), 2) the sensory network (S1, S2, mINS, and pINS), and 3) the affective network (dlPFC, sgACC, and aINS). The CONN-fMRI toolbox was used to test for a statistically significant difference in RSFC between males and females (P < 0.05). Because our ROI-to-ROI a priori hypotheses were developed at the network level in which specific ROIs show an effect, we report unadjusted P values, as suggested by the developers of Conn. “Uncorrected P values are appropriate when the researcher’s original hypotheses involve only the connectivity between two a priori ROIs and false discovery rate (FDR)–corrected P values are appropriate when the researcher’s original hypotheses involve the connectivity between larger sets of ROIs and do not specify a priori which ROIs are expected to show an effect” [47]. See Supplementary Table 2 for both uncorrected and FDR corrected values.

Analysis of Head Motion

The threshold for the maximum motion was set to 1 mm. Artifact Detection Toolbox (ART; www.nitrc.org/projects/artifact_detect/) was used to identify individual volumes contributing significant noise (z = 9, threshold = 1 mm), after which the outlying volumes were regressed and removed at subject-level analysis. No greater than 5% of fMRI dynamics for any given subject was excluded from the analysis. The median motion (in mm) for all participants in this analysis was 0.25 mm (min = 0.10 mm, max = 0.72 mm) and was not statistically significantly different by group (median: males = 0.24 mm, females = 0.25 mm, P = 0.651).

General and Psychophysical Analyses

Demographic and all study variables were summarized for the entire sample and each sex group using descriptive statistics. Frequency distributions summarized nominal and ordinal categories. Due to skewness, continuous data (e.g., age, MMSE, psychophysics) were summarized using median and interquartile range (IQR). Group comparisons were conducted via SPSS (version 23) using chi-square tests of independence (nominal, ordinal data) and Mann-Whitney Tests (continuous data). Psychophysics and RSFC data were rank-transformed as necessary to meet normality assumptions of Pearson correlations. Those correlations were used to assess the strength of the associations between the psychophysical reports and the extracted RSFCs from the ROI analyses. We used z tests of independent correlations to test for differences in the direction and strength of those associations between males and females. An unadjusted alpha P value of less than 0.05 was used for determining statistical significance for this analysis. SPSS (version 23) was used for these analyses.

Results

Demographic and Psychophysical Reports

As shown in Table 1, there were no statistically significant differences in demographic or psychological characteristics between males and females (all P > 0.13). There were no statistically significant differences between males and females in temperature thresholds or unpleasantness ratings of the three targeted perceptual intensities (all P > 0.23) (Table 2). Effect sizes (Cohen’s d) for the differences between the temperature thresholds at mild and moderate pain were 0.30 and 0.27 for males and females. Cohen’s d for the difference in affective ratings at the moderate pain intensity was 0.38 (Table 2).

Table 1.

Demographic and clinical summaries by sex

	Total (N = 38)	Female (N = 19)	Male (N = 19)	P
	No. (%)	No. (%)	No. (%)
Race				0.513
African American	3 (7.9)	2 (10.5)	1 (5.3)
Asian	1 (2.6)	0 (0.0)	1 (5.3)
Caucasian	34 (89.5)	17 (89.5)	17 (89.5)
Education	N = 37	N = 18	N = 19	0.318
High school or less	4 (10.8)	3 (16.7)	1 (5.3)
Partial college	10 (27.0)	4 (22.2)	6 (31.6)
College grad	9 (24.3)	6 (33.3)	3 (15.8)
Grad school	14 (37.8)	5 (27.8)	9 (47.4)
Marital status				0.238
Single/divorced/separated	8 (21.1)	5 (26.3)	3 (15.8)
Married	23 (60.5)	9 (47.4)	14 (73.7)
Widowed	7 (18.4)	5 (26.3)	2 (10.5)
	Median (IQR)	Median (IQR)	Median (IQR)
Age, y	68.5 (66–81)	70.0 (66–80)	68.0 (66–81)	0.883
Total SES score	57.5 (45–65)	58.0 (53–64)	57.0 (45–66)	0.965
MMSE score	30.0 (29–30)	30.0 (28–30)	30.0 (29–30)	0.987
BPI-SF average pain	1.0 (0–2)	0.0 (0–2)	1.0 (0–2)	0.914
BPI-SF pain right now	0.0 (0–1)	0.0 (0–0)	0.0 (0–1)	0.156
GDS-SF score	0.0 (0–1)	0.0 (0–1)	0.0 (0–1)	0.485
STAI state score	48.0 (45–51)	50.0 (45–53)	47.0 (45–50)	0.139
STAI trait score	47.0 (44–50)	47.0 (44–50)	47.0 (45–49)	0.860

BPI-SF = Brief Pain Inventory Short Form (range = 0 no pain to 10 most pain); GDS-SF = Geriatric Depression Scale Short Form (range 0 = no indication of depression to 15 high possibility of depression); MMSE = Folstein Mini Mental State Examination (range = 0 completely cognitively impaired to 30 completely cognitively intact); SES = Hollingshead Four Factor Index of Social Status (range = 8 lowest to 66 highest); STAI = Spielberger State or Trait Anxiety Inventory (range = 20 increased anxiety to 80 least amount of anxiety).

Table 2.

Summary of psychophysics of temperature thresholds necessary to produce warmth, mild pain, or moderate pain and unpleasantness ratings at each condition (N = 38, 19 male and 19 female)

Variables		Min	Max	Median	IQR	P	Effect size*
Temperature
Warmth	Males	31	38	32	32.0–33	0.469	0.23
	Females	31	36	32	32.0–33
Mild pain	Males	33	47	37	34.0–41	0.295	0.30
	Females	34	38	35	34.0–38
Moderate pain	Males	37	48	42	37–45	0.490	0.27
	Females	36	46	39	38–42
Unpleasantness
Warmth	Males	0	8	0	0–2	0.357	0.27
	Females	0	3	0	0–1
Mild pain	Males	0	16	1	0–5	0.700	0.17
	Females	0	6	2	0–5
Moderate pain	Males	2	19	6	5–11	0.230	0.38
	Females	0	11	5	4–8

P values derived from Mann-Whitney test.

IQR = interquartile range.

^*Cohen’s d from transformed normal data.

Resting State Functional Connectivity

Six statistically significantly different ROI-to-ROI pairs were identified in the three networks. Relative to females, males showed greater RSFC between the right (R-) dlPFC and R-mINS (beta = 0.19, P = 0.009), between the left (L-) dlPFC and R-S2 (beta = 0.18, P = 0.032), and L-sgACC and L-AMY (beta = 0.09, P = 0.041). Conversely, females showed greater RSFC between L-HYPO and L-S1 (beta = 0.16, P = 0.012), L-HYPO and L-S2 (beta = 0.14, P = 0.014), and R-sgACC and PAG (beta = 0.13, P = 0.017). Figure 1 shows significantly increased and decreased RSFC in ROI-to-ROI pairs in our conceptual model.

Figure 1.

Results of sex differences in resting state functional connectivity (RSFC) between ROI to ROI. Circle (orange) regions compose part of the affective pain network. Square (green) regions compose part of the sensory pain network. Oval (blue) regions compose part of the descending modulatory network. Increased RSFC in males is depicted by wide (blue) lines, and increased RSFC in females is depicted by narrow (orange) lines. RSFC = resting state functional connectivity; ROI = region of interest.

Association Between Psychophysics and Resting State Functional Connectivity

Tests of differences between the groups in the strength and direction of the associations between psychophysics and RSFC were conducted using z tests of independent correlations (Table 3). The strongest evidence for sex differences emerged for the associations of thermal percepts with RSFC between L-sgACC and L-AMY and for RSFC between L-HYPO and L-S2. A statistically significantly positive correlation of the RSFC between L-sgACC and L-AMY and the moderate pain temperature threshold was observed for females (r = 0.77, P < 0.001) but not males (r = 0.04, P = 0.888; z = 2.77, P = 0.006). RSFC between L-HYPO and L-S2 was significantly and inversely associated with mild pain temperature thresholds for females (r = –0.49, P = 0.034), with a considerably lower positive correlation observed for mild pain in males (r = 0.22, P = 0.377; z = 2.15, P = 0.032). Weaker evidence of sex differences was observed for associations between RSFC between L-HYPO and L-S1 and temperature threshold for perceptions of warmth. A statistically significant inverse correlation was found in females (r = –0.48, P = 0.040) but not males (r = 0.09, P = 0.718), yet the difference between the groups only approached statistical significance (z = 1.73, P = 0.084).

Table 3.

Associations of resting state connectivity with psychophysics by sex

	Temperature			Affective
Temperature	Female	Male	Difference P	Female	Male	Difference P
	(N = 19)	(N = 19)		(N = 19)	(N = 19)
Right dlPFC to right middle insula
Warmth	0.19	−0.33	0.131	−0.15	−0.08	0.842
	(0.448)	(0.170)		(0.547)	(0.747)
Mild pain	−0.13	−0.04	0.795	−0.18	−0.44	0.206
	(0.609)	(0.859)		(0.475)	(0.059)
Moderate pain	−0.14	−0.01	0.711	0.11	−0.37	0.159
	(0.571)	(0.958)		(0.669)	(0.124)
Left dlPFC to right S2
Warmth	0.12	−0.42	0.107	0.001	−0.04	0.905
	(0.619)	(0.071)		(0.998)	(0.859)
Mild pain	−0.24	0.17	0.238	−0.24	−0.42	0.569
	(0.316)	(0.499)		(0.318)	(0.071)
Moderate pain	−0.32	0.03	0.308	0.30	−0.26	0.103
	(0.188)	(0.904)		(0.215)	(0.286)
Left hypothalamus to left S1
Warmth	−0.48	0.09	0.084	0.26	−0.03	0.401
	(0.040)	(0.718)		(0.282)	(0.916)
Mild pain	−0.40	0.09	0.147	0.25	0.23	0.952
	(0.089)	(0.731)		(0.304)	(0.336)
Moderate pain	−0.35	−0.10	0.219	−0.04	0.16	0.569
	(0.147)	(0.692)		(0.861)	(0.526)
Left hypothalamus to left S2
Warmth	−0.40	−0.23	0.589	−0.15	0.13	0.424
	(0.194)	(0.341)		(0.546)	(0.586)
Mild pain	−0.49	0.22	0.032	−0.06	−0.20	0.689
	(0.034)	(0.377)		(0.808)	(0.414)
Moderate pain	−0.49	0.01	0.124	−0.17	−0.08	0.795
	(0.035)	(0.953)		(0.489)	(0.740)
Left sgACC to left amygdala
Warmth	0.31	−0.11	0.223	−0.15	0.46	0.067
	(0.194)	(0.651)		(0.530)	(0.049)
Mild pain	0.56	0.07	0.112	0.19	0.08	0.749
	(0.012)	(0.778)		(0.426)	(0.733)
Moderate pain	0.77	0.04	0.006	0.31	−0.04	0.308
	(< 0.001)	(0.888)		(0.197)	(0.872)
PAG to right sgACC
Warmth	0.25	0.01	0.490	0.01	−0.23	0.490
	(0.307)	(0.962)		(0.979)	(0.336)
Mild pain	0.10	−0.23	0.342	−0.05	−0.04	0.976
	(0.673)	(0.337)		(0.847)	(0.882)
Moderate pain	−0.10	−0.11	0.976	−0.06	−0.09	0.928
	(0.686)	(0.670)		(0.822)	(0.725)

Values in cells: correlation (P value).

dlPFC = dorsolateral prefrontal cortex; PAG = periaqueductal gray matter; sgACC = subgenual anterior cingulate cortex.

Regarding associations of the RSFCs with unpleasantness reports, there was a statistically significantly positive correlation of the RSFC between L-sgACC and L-AMY and unpleasantness in the group of males (r = 0.46, P = 0.049). If anything, that association was inverse in the females (r = –0.15, P = 0.531), but the difference between the groups did not reach the criterion for statistical significance (z = 1.83, P = 0.067).

Discussion

The current study examined sex differences in brain regions modulating pain by examining the psychophysical response to experimental thermal pain, the RSFC of brain regions implicated in pain processing, and the associations between psychophysical responses and RSFC. Our first hypothesis (psychophysical) was that when compared with older males, older females would report increased pain sensitivity and greater unpleasantness. This hypothesis was not supported. Unlike the previously published pilot sample in which we found that older females were more sensitive to evoked thermal pain, that finding was not borne out in this larger sample. However, as in the pilot sample, we did not find significant sex differences in the ratings of unpleasantness.

While counter to our hypothesis, these findings are not particularly surprising as the magnitude of sex differences in heat pain perception has been quite variable across studies. For pain intensity, our results are in the direction of greater pain sensitivity in females, but the magnitude of the difference was small. This is consistent with a previous meta-analysis, which showed small to moderate effect sizes for sex differences in heat pain threshold and tolerance [49]. For pain unpleasantness, our findings are in the direction of greater pain unpleasantness in males, but the magnitude of the differences was also small and in the opposite direction of most previously published studies [5].

Our second hypothesis (neurophysiological) was that RSFC between sensory, affective, and modulatory pain regions would differ by sex. We predicted that when compared with older males, older females would demonstrate increased RSFC between the three regions forming the affective pain network (dlPFC, sgACC, aINS) and the four regions comprising the sensory pain network (S1, S2, mINS, pINS). We further posited that when compared with older females, older males would demonstrate greater RSFC between the four regions in the sensory pain network (S1, S2, mINS, pINS) and the three regions in the descending modulatory network (AMY, HYPO, and PAG). This hypothesis was partially supported. We found that older women had greater RSFC between components of the sensory and descending modulatory networks (S1 and L-HYPO, S2 and L-HYPO) and between elements of the affective and descending modulatory networks (R-sgACC and PAG). While we did find that older males had greater RSFC between the affective (sgACC) and descending modulatory networks (L-AMY), we did not find that older females had a more significant RSFC between the sensory network and the affective network.

Our third hypothesis was that RSFC would correlate with psychophysical responses. This hypothesis was partially supported. While we did not find associations between all RSFC and psychophysical responses, we did find sex differences in the association of L-HYPO and L-S2 with mild thermal pain and the association of L-sgACC and L-AMY with moderate thermal pain. These differences were driven by a strong negative association of L-HYPO and L-S2 RSFC with mild pain in females and a strong positive association of L-sgACC and L-AMY RSFC with moderate pain in females.

Our findings indicate potential sex differences in pain modulatory systems in older adults, and several of our results support and extend recent findings from a sex differences study examining RSFC in healthy younger adults [23]. In the current study, we decided to include the sgACC as an important component of the affective network [23]. Based on previous psychophysical findings demonstrating that, relative to men, younger women show greater habituation to heat pain [22,49–51], Wang and colleagues decided to examine sex differences in RSFC and structural connectivity in pain modulation in a sample of young adults age 18 to 36 years [23]. They found that women had stronger RSFC between the sgACC and the descending modulatory system. They found that when compared with males, females had greater RSFC between sgACC and PAG, raphe nucleus, medial thalamus, and anterior midcingulate cortex. Conversely, when compared with females, males exhibited greater RSFC between the aINS and temporal-parietal junction and between sgACC and HYPO. Wang and colleagues concluded that brain circuitry in women seems to provide stronger “engagement” of the descending modulatory system and that men better “engage” a system that attends to pain, which would prevent habituation. A conclusion was that hypothalamus findings in men indicate a greater pain-evoked stress response when compared with women.

Consistent with Wang and colleague’s results [23], we found that, relative to older males, older females have stronger RSFC between the R-sgACC and PAG (representing pain affect and descending modulation, respectively). Unlike Wang and colleagues, we included the HYPO as part of the descending modulatory network as described by Tracey and Mantyh [11], and we found greater RSFC between S1 and L-HYPO and between S2 and L-HYPO among women (reflecting greater links between sensory pain systems and descending modulation). In contrast to Wang and colleagues’ findings [23], these results could also possibly indicate a greater stress response in females relative to males. Our findings in older males are not as strongly aligned with Wang and colleagues. However, in one area there seems to be partial agreement. They found that younger males had increased RSFC between sgACC and HYPO and hypothesized that this indicated a greater stress response in males. We found that, relative to older females, older males had greater RSFC between sgACC and L-AMY, which may likewise be related to a greater pain-related stress response among males.

The strongest overlap and extension of Wang and colleagues’ work in young adults is that both their findings and the current findings in older adults are that, relative to males, females exhibit greater FC between the sgACC and PAG. The PAG is well known to be the epicenter of the descending modulatory system, particularly the endogenous opioid component of that system [44].

Limitations

There are some limitations in the current study. Because the current study was part of a large-scale study with multiple scanning sequences, the resting state scan acquisition occurred after 16 minutes of a functional pain-task paradigm. However, a full minute of rest occurred before the resting state sequence began. Thus, the research team determined that there would be no issue with residual activation from the previous paradigm, though we could not guarantee that. Also, the use of different methods for assessing pain intensity vs pain unpleasantness could have influenced the results. Finally, our inclusion and exclusion criteria led to an unusually healthy group of older adults, with higher SES, which could limit the generalizability of the findings to the larger older adult population.

Conclusions

The current study results hint that, relative to older males, older females demonstrate stronger associations between sensitivity to evoked thermal pain and the degree of RSFC between descending modulation and both the sensory network (L-S2 and L-HYPO) and the affective network (L-sgACC and L-AMY). Taken together, the current findings and findings from studies of sex differences in younger populations describe a possible phenomenon in which younger and older females may exhibit increased RSFC between regions typically associated with the sensory and affective components of pain and the descending modulatory system. This supposition seems to support the recent meta-analyses demonstrating that females require less morphine to achieve similar analgesic responses to males. A long-standing hypothesis is that, when compared with males, females may be at greater risk of experiencing greater pain sensitivity. However, a pattern seems to be emerging that females may better engage the descending modulatory network, thus potentially protecting females from suffering from pain unpleasantness while males may be at risk of suffering more pain unpleasantness. Future studies include repeating the current methods to potentially confirm findings in a larger sample composed of a wider range of adults (age 30–89 years). To further describe sex differences, we plan to examine task-evoked functional connectivity in the current data set. It is important that future studies include both healthy adults and older adults with chronic pain.

Authors’ Contributions

TBM contributed to the study conception, design, data analysis, data collection, and interpretation of data. RBF contributed to the data analysis and interpretation. SPB contributed to the study conception, design, and interpretation of data. BPR contributed to the study conception, design, and interpretation of data. MSD contributed to the study conception, design, and interpretation of data. JCG contributed to the study conception, design, and interpretation of data. SWA contributed to the study conception, design, data collection, analysis, project management, and data interpretation. RLC contributed to the study conception, design, data analysis, and interpretation of data. All authors approved the final version of the manuscript.

Funding sources: This work was supported by the John A. Hartford Foundation, the Mayday Fund, the Vanderbilt Office of Clinical and Translational Scientist Development, the Vanderbilt Institute of Clinical and Translational Research (grant V4007), the Vanderbilt Clinical and Translational Research Scholars Program, and the National Institutes of Health National Institute on Aging (grants K23 AG046379-01A1 and R21 AG045735-02). The contents are solely the responsibility of the authors and do not necessarily represent the official views of these institutions.

Study data were collected and managed using the REDCap electronic data capture tools hosted at Vanderbilt University. REDCap is maintained by Vanderbilt Institute for Clinical and Translational Research, which is supported by the National Institutes of Health National Center for Advancing Translational Sciences (grant L1 TR000011).

Conflicts of interest: The authors have no conflicts to disclose.