Functional connectivity associations with menstrual pain characteristics in adolescents: an investigation of the triple network model.

Participants were recruited from social media advertisements, a posting on the institution’s centralized research recruitment website, and word of mouth. Inclusion criteria included: 1) females ages 13-19 years; 2) self-reported menstrual cycle averaging 22-35 days; 3) regular menstrual cycles for at least 6 months; 4) right-handed; 5) body mass index (BMI) of 35 or less; 6) able to read and understand English; 7) access to a smartphone or email. Exclusion criteria included: 1) use of oral contraceptives or any exogenous hormones in the previous 3 months prior to participation; 2) presence of factors indicative of secondary dysmenorrhea (e.g., self-reported presence of persistent pelvic pain throughout the month); 3) diagnosis of chronic pain condition (e.g., irritable bowel syndrome (IBS), functional abdominal pain, chronic migraines, interstitial cystitis/painful bladder syndrome); 4) current self-reported severe depression, bipolar disorder, panic disorder, or ADHD, or current treatment for these conditions; 5) diagnosis of an eating disorder within the last 6 months; 6) current or past diagnosis of any psychotic disorder; 7) currently pregnant or breastfeeding; 8) self-reported weekly use of alcohol, cannabis, and/or other illegal substances; 9) Use of stimulants (including methamphetamine and/or medications for the treatment of ADHD) or opioids in the previous 3 months. Participants who use other analgesics will be included but will be requested to not take these analgesics within the previous 24 hours of the laboratory session; 10) history of pelvic inflammatory disease or sexually transmitted disease; 11) acute illness or injury that would potentially impact pain task performance (e.g., fever, flu symptoms) or that affect sensitivity of the extremities (e.g., Reynaud’s disease); 12) developmental delay, diagnosis of autism, or significant cognitive impairment that may preclude understanding of study procedures; 13) presence of certain ferromagnetic appliance or implants (braces, retainers, spacers, wires, screws, etc.) in the mouth or any other body part, which is a contraindication for the magnetic resonance imaging (MRI) scanner; 14) significant claustrophobia. Two thousand four hundred and forty-seven individuals expressed interest in participating. Of these individuals, many were ineligible, unable to be contacted, or were no longer interested (see Figure 1 for a detailed flowchart of participant recruitment). One hundred and three participants completed the study visit. Study visits occurred during the mid-follicular phase (days 8-14 of the menstrual cycle), which was determined based on the most recent first day of menstruation assessed at the Intake visit. Study visits included the MRI session and self-report measures of menstrual pain characteristics as described below. Analyses presented are for the first study visit of a larger, ongoing longitudinal study. Participants are compensated for each study component they complete, up to $530 for completion of the entire study. The study was approved by the Mass General Brigham IRB under protocol 2019P001729. All minor participants (ages 13-17) and a legal guardian provided written informed assent and parental permission; young adult participants (ages 18-19) provided written informed consent. Demographic variables (e.g., date of birth, race, and ethnicity) and menstrual history variables (age at menarche, menstrual pain rating, etc.) were assessed using an instrument designed for this study. Usual level of menstrual pain without pain medication was assessed using an 11-point numeric rating scale (NRS) from 0 (no pain) to 10 (worst pain possible) [ 6 ; 14 ]. To assess menstrual pain interference, participants used the same NRS (with anchors at “not at all” and “interferes completely”) to answer the question: “How much does pain during your period interfere with your daily life (for example: school, work, physical and social activities, relationships, sleep, etc.)?” Participants who reported no menstrual pain (i.e., 0 on the first NRS), were not asked the pain interference question and values of 0 were imputed. Cumulative menstrual pain exposure was assessed with a 5-point categorical question: “Over the course of your life, how many menstrual periods have you experienced pain?” Answer choices included: none, 1 or 2 periods, 3-6 periods, 7-12 periods, 13-24 periods, and more than 24 periods. Blood oxygenation level dependent (BOLD) eyes-open resting-state fMRI data (TR=750ms, TE=25ms, flip angle=52, 72 slices, field of view=220 mm, voxel size 2x2x2mm, multiband acceleration factor=6, total duration=6.0 min, total volumes=480) and a high-resolution multi-echo T1-weighted MPRAGE anatomical image (TR=2530ms, TEs=1.69ms, 3.55ms, 5.41ms and 7.27ms, flip angle=7, 176 slices, field of view=256mm, voxel size 1x1x1mm) were collected using a 3T Prisma scanner (Siemens, Inc) and a 64-channel phased-array head coil. During the resting state scan, participants were instructed to lie still with their eyes open and focus on a fixation cross being displayed via projector. During the T1 scan, participants were instructed to watch a neutral video that has been shown to effectively reduce head motion (Inscapes with no scanner sound; [ 57 ]). Structural and functional MRI data were organized into Brain Imaging Data Structure (BIDS) format [ 23 ]. Resting state and structural data were then preprocessed using fMRIPrep 20.2.1 [ 17 ; 18 ], which is based on Nipype 1.5.1 [ 21 ; 22 ]. Detailed pre-processing information provided in the fmriprep reports, which are released under CO licence, are provided in the supplementary material . We briefly summarize preprocessing steps as follows: Anatomical data preprocessing: The T1-weighted (T1w) image was corrected for intensity non-uniformity, skull-stripped and segmented into cerebrospinal fluid (CSF), white-matter (WM) and gray-matter (GM) using FSL FAST [ 66 ]. Volume-based spatial normalization was then performed through nonlinear registration with antsRegistration (ANTs 2.3.3). Functional data preprocessing: The following preprocessing was performed: susceptibility distortion correction [ 31 ; 61 ], registration to the T1w reference using bbregister (FreeSurfer) which implements boundary-based registration [ 26 ], motion correction (FSL 5.0.9, [ 32 ]), slice-time correction (3dTshift from AFNI 20160207 [ 9 ], and registration to standard space. Automatic removal of motion artifacts using independent component analysis (ICA-AROMA, [ 49 ]), was performed on the preprocessed BOLD MNI space time-series after spatial smoothing with an isotropic, Gaussian kernel of 6mm FWHM (full-width half-maximum). Corresponding “non-aggressively” denoised runs were produced after ICA-AROMA. Quality assurance measures of fractional displacement (FD; [ 32 ; 48 ]) and DVARS (temporal derivative of root mean square variance over voxels; [ 47 ]) were computed from the denoised data. Finally, the first 10 non-steady state volumes were removed from the denoised fMRI data and temporal highpass (150sec) filtering and masking using MNI6AsymNLin mask to exclude non-brain regions were applied before running further analyses. Volumes that exceeded a threshold of 0.5 mm FD or 1.5 standardised DVARS were annotated as motion outliers by fmriprep. Participants were excluded for excessive head motion if either the % of volumes classified as motion outliers exceeded 10% or if the number of volumes with FD > 2mm exceeded 5. Estimating Resting State Networks (RSNs): Group Independent Component Analysis (GICA) with different model orders (30, 35, 40 and 45) was implemented using FSL MELODIC [ 4 ; 5 ] to identify resting state functional connectivity networks. These model orders were chosen because they have been consistently shown to reveal large scale networks in the group GICA results, with the specific model order varying around this range from dataset to dataset due to dataset-specific noise [ 4 ; 43 ]. GICA spatial maps were visually inspected by authors PK and LN to determine the optimal model order that identifies sets of independent components that most closely resembled previously reported resting state networks [ 34 ; 43 ; 55 ]. Using this approach, we selected model order 35 as the parcellation scheme for further analyses. Selection of a priori RSNs: Four components corresponding to the Triple Network Model [TNM; 42 ] were identified from the GICA maps for investigating the association of our variables of interest with connectivity of these networks: Default Mode Network (DMN), left and right Central Executive Networks (l/rCEN) and Cingulo-Opercular Salience Network (cSN). Dual Regression: The GICA maps were then thresholded at a z score of +/−2 and normalized to a maximum value of 1 by dividing each map by its maximum z score to account for differences in the scale of the spatial maps. Subject-specific network maps that encode the individual variability in functional connectivity were obtained via dual regression [ 44 ]. First, subject-specific time courses for each GICA component were extracted simultaneously via multivariate spatial regression of the scaled GICA spatial maps against each subject’s fMRI timeseries. Then these network time courses were regressed against each participant’s fMRI data to estimate subject-specific spatial maps corresponding to each GICA component. These maps are comprised of regression coefficients at each voxel that denotes the functional connectivity of each voxel with the corresponding resting-state network represented by the GICA component. To investigate the association of menstrual pain severity and interference, and cumulative menstrual pain exposure on the TNM networks, we implemented 12 general linear models with subject-specific spatial maps for each of the four a priori TNM networks as dependent variables and pain measures as independent variables (three measures analyzed separately), and chronological age and gynecological age (number of years menstruating) as covariates of no interest. Each model was evaluated using FSL Randomize for non-parametric permutation testing to produce a cluster mass statistic image with cluster forming threshold of z=1.6 and inference using 5000 permutations to control family-wise error (p<0.05). The cluster mass statistic has been shown to be more sensitive than cluster size [ 8 ; 30 ] and non-parametric permutation testing is valid at any threshold and degrees of freedom with sufficient smoothness (> 3mm) [ 30 ].

Of the 103 participants who completed the study visit, two participants were excluded due to excessive motion based on the QA criteria and one participant was excluded due to artifacts in the raw MRI data. Therefore, the final sample included 100 participants. Demographic variables are presented in Table 1 . All three menstrual pain measures were associated with connectivity within or between at least one of the TNM networks. Both cSN and left CEN were associated with menstrual pain severity. Specifically, menstrual pain was positively associated with functional connectivity between the cSN and both of nodes within cSN and with the auditory cortex, limbic regions including the amygdala and hippocampus, and insula ( Figure 2A ). Menstrual pain was also positively associated with connectivity between the left CEN and posterior regions including the precuneus, posterior cingulate, and parietal cortices ( Figure 2B ). Connectivity of cSN with widespread brain areas that overlap with other large scale brain networks such as DMN, executive function, and CEN was stronger in individuals with greater pain interference ( Figure 3A ). In addition, we found that pain interference was also associated with connectivity between cSN and PAG and sensorimotor cortex. In addition, both left and right CEN were associated with pain interference, however in opposite directions. Specifically, connectivity within the left CEN was associated positively with pain interference, whereas the connectivity of the right CEN, both of nodes within right CEN and with other cortical areas including the insula outside the network, were negatively associated with pain interference ( Figure 3B ). Connectivity of the DMN was strongly associated with cumulative menstrual pain exposure, with cumulative menstrual pain exposure showing a strong negative association with connectivity between widespread brain areas within the DMN and other regions including areas overlapping other large-scale brain networks, such as the cSN ( Figure 4 ).

The present study aimed to identify the associations between the connectivity of three brain networks specified in the TNM of psychopathology – CEN, cSN, and DMN – and variables assessing different dimensions of menstrual pain in a sample of adolescent girls with varying levels of menstrual pain associated with PD. This is the first study to investigate the relationship of neural alterations in young girls with PD and extends beyond previous research in adult women with menstrual pain to focus on the dimensional analyses of the menstrual pain variables, without grouping participants based on an arbitrary cutoff measure of self-reported pain severity. Additionally, we controlled for chronological age and gynecological age in our analyses, suggesting our results are related to the experience of pain, rather than developmental changes over the course of puberty. Results indicated significant relationships between all three menstrual pain variables (menstrual pain severity, menstrual pain interference, and cumulative menstrual pain exposure) and the functional connectivity of key cognitive-emotional networks associated with pain [ 33 ]. Menstrual pain severity showed a strong positive correlation with functional connectivity in the cSN, suggesting that as menstrual pain severity increases, the connectivity within cSN and between cSN and other brain regions also increases. The cSN is a network that is associated with personal relevance and the meaning of external and internal experiences [ 42 ]; as such, increases in menstrual pain severity may be associated with an increased awareness of somatic experiences. This notion is also supported by the strong, positive relationship of cSN connectivity and menstrual pain interference. Interestingly, we also found patterns of connectivity with pain interference and the cSN and the sensorimotor cortex and the PAG – a finding previously reported by Wei et al. [ 62 ]. However, our results suggest that these alterations in connectivity are present even in the non-painful phases of the menstrual cycle, supporting the idea that connectivity patterns reflect stable changes rather than reflecting the experience of being in pain during menstruation. Interestingly, the CEN, which is often associated with functional impairment in individuals with chronic pain [ 16 ], also demonstrated significant, positive associations between the functional connectivity within this network and ratings of menstrual pain severity and pain interference. Specifically, the CEN plays a role in critical cognitive functions, including executive functions and cognitive regulation of internal events and external stimuli [ 56 ]. Our findings that functional connectivity of the left and right CEN each share overlapping and distinct relationships with menstrual pain and menstrual pain interference suggests that cognitive evaluation and regulation of internal experiences are important in adolescent girls with PD, particularly for those with higher levels of menstrual pain severity and interference. For example, past research has highlighted the role of pain catastrophizing in menstrual pain [ 19 ; 36 ; 60 ]. Pain catastrophizing is a term used to describe a maladaptive and negative cognitive style associated with exaggerated responses to and fears of pain [ 50 ]. This is in line with our data demonstrating that areas of the brain involved in cognitive evaluation and regulation of pain may be a mechanism involved in menstrual pain or reflect a neural phenotype of girls at risk for chronic pain. Future research identifying neural correlates of cognitive processes in menstrual pain is needed to further identify the ways that cognitive functions interact with menstrual pain. At rest, the DMN is often engaged in idle thinking, autobiographical recall, Theory of Mind processing, and self-reflection, and in the current study, connectivity between the DMN and widespread brain regions was negatively correlated with cumulative menstrual pain exposure. Although we cannot make specific clinical inferences based on these data, one interpretation is that the extent to which an individual has experienced menstrual pain over the course of her life is associated with changes in brain connectivity of a key network involved in everyday tasks and overall functioning. Key nodes of the DMN, such as the posterior cingulate cortex, have been implicated in higher level processes involved in chronic pain in adult fibromyalgia patients [ 16 ]. Additionally, De Ridder argues that the experience of pain and suffering leads the DMN to become pathologically “coupled” with the somatosensory network, thereby resulting in the adoption of pain as part of one’s self-identity [ 12 ]. The results from the present study support this notion, as the strength of DMN functional connectivity is related to the number of menstrual cycles with menstrual pain, even in adolescents. Thus, our data further support the role of the DMN in the chronification of pain. While previous studies have demonstrated alterations in women with PD (compared to women without PD) in the DMN [ 41 ; 62 ], as well as other brain regions consistent with neural signatures evident in individuals with chronic pain, such as the amygdala [ 54 ], anterior cingulate cortex [ 39 ; 40 ; 65 ], insula [ 13 ; 63 ], thalamus [ 28 ], and periaqueductal gray [ 62 ], we are not aware of any studies that have examined these processes in adolescent populations with PD. Only one imaging study has specifically looked at young adult women with PD (mean age of 23 years) and found no differences in global network metrics or modular functional architecture between individual brain areas (using an anatomical parcellation) compared to healthy young women [ 35 ]. The results from the current study directly address this significant gap in the literature and suggest there are stable neural changes in pain processing in adolescent girls with PD without chronic pain. This offers the possibility that adolescent girls with menstrual pain may be associated with pre-existing neural vulnerabilities to experiencing pain; however, the relationship of cumulative menstrual pain exposure to alterations in the DMN also suggests that this is a dynamic process that changes over time and with repeated experiences of pain. Future research should continue to explore neural processing during the emergence of menstrual pain in adolescents. There are some limitations that should be considered when interpreting our findings. First, these data were obtained during resting state during a pain-free phase of the menstrual cycle. As such, we are unable to determine alterations in pain processing during the actual experience of pain, which may reflect different brain regions or patterns of connectivity. Second, all girls participating in this study had been menstruating regularly for at least six months prior to enrollment, so we cannot be sure whether these changes would be evident even earlier in the pubertal phase of development. Additionally, participants in this study did not receive a laparoscopy confirming or disconfirming the diagnosis of PD. It is possible that some participants in this cohort will ultimately be diagnosed with a pelvic condition, such as endometriosis. However, we do not anticipate that the findings in the current study would be different based on whether participants have a diagnosis of PD or secondary dysmenorrhea, based on literature showing a poor correlation between degree of endometriosis and pain experience [e.g., 10 ; 27 ]. Further, we chose not to correct for multiple comparisons (number of pain measures and networks tested) as our approach, which is a blend of data-driven statistical analysis combined with hypothesis testing, included stringent corrections for the number of voxels tested for the whole brain associations that were examined. Not further correcting for number of pain measures and networks allows us to maintain some protection against false negatives while still emphasizing protections against false positives. We note that if a Bonferroni correction is applied, the findings of associations with pain measures and connectivity of CEN and DMN do not reach a corrected level of significance, while findings with cSN remain significant at the more stringent threshold. As such, our findings may indicate underlying patterns that warrant further investigation with larger sample sizes or different methodologies. Overall, this study is the first of its kind to examine associations between brain network functional connectivity and different characteristics of menstrual pain (menstrual pain severity, menstrual pain interference, and cumulative menstrual pain exposure) in adolescent girls with a range of menstrual pain severity. Importantly, results indicate that there are brain-based alterations in connectivity associated with each of the variables, suggesting stable changes in how the brain processes pain-related information and experiences in adolescent girls. Our findings highlight the importance of examining pain experiences in adolescents and support the need for additional research identifying how these alterations may or may not change over time and whether they are associated with other risk factors for chronic pain.

Menstrual pain is a common and disabling condition resulting in substantial loss of productivity in about 25% of reproductive age girls and women [ 2 ; 24 ; 52 ] and is classified as either primary dysmenorrhea (PD; menstrual pain without any identified pathology) or secondary dysmenorrhea (menstrual pain associated with another condition such as endometriosis, adenomyosis, etc.). Adolescent girls report significant interference in school functioning, with menstrual pain causing both absenteeism and difficulty performing in class [ 1 ; 51 ]. Importantly, menstrual pain in adolescence is also a risk factor for future chronic pain conditions [ 20 ; 29 ; 37 ]. Recent research has suggested that central sensitization may play a role in pain severity in PD, e.g., [ 46 ; 59 ]. Evidence for central sensitization [ 64 ] comes from numerous studies demonstrating heightened pain sensitivity across pain testing paradigms and body locations in women with PD compared to women without PD [ 45 ]. Data demonstrating abnormalities in activation of brain areas in adult women with PD has also supported the notion that menstrual pain may be related to alterations in pain processing [ 35 ; 38 – 41 ; 62 ]. One of the first studies to examine functional magnetic resonance imaging (fMRI) in response to thermal pain in women with and without PD found widespread deactivation of brain regions, especially in the default mode network (DMN) in women without PD that was not observed in women with PD. The DMN is a key neurocognitive brain network comprised of medial prefrontal cortex, posterior cingulate, temporal areas, and parietal cortex that is active during internal mentalizing.[ 59 ]. The authors suggest that these data may be evidence of alterations in networks that are active at rest in women with PD. Interestingly, similar patterns of impaired DMN deactivation have also been demonstrated in individuals with chronic low back pain [ 3 ]. Yet, existing studies all suffer from a common limitation – the classification and grouping of study participants into “PD” or “healthy control” based on a single rating of severity of menstrual pain, with an arbitrary cutoff score to classify those with PD, which may not represent a clinically meaningful difference [ 15 ]. The present study sought to explore potential neural mechanisms involved in PD, primarily by focusing on an adolescent population and analyzing menstrual pain characteristics as continuous variables. We focused on three primary cognitive-emotional networks (DMN, the cingulo-opercular salience network (cSN), and the central executive network (CEN)) of the triple network model (TNM) of psychopathology as described by Menon [ 42 ], which was recently adopted by De Ridder and colleagues as a unifying framework to investigate pain pathways [ 11 ; 12 ]. These networks include the self-representational DMN [ 7 ; 25 ], the behavioral relevance encoding cSN (including insular and dorsolateral anterior cingulate cortices) [ 53 ], and the goal-oriented CEN (including lateral, frontal, and posterior parietal cortices) [ 53 ; 58 ], which reflect key neurocognitive networks that may play a role in aspects of menstrual pain, such as cognitive functioning and attention [ 11 ; 12 ]. We hypothesized that functional connectivity within and between these networks would be associated with all three menstrual pain characteristics: severity, interference, and cumulative menstrual pain exposure.