Within-day rhythms of pain and cognitive function in people with and without fibromyalgia: Synchronous or syncopated?

Introduction

Up to 90% of people with fibromyalgia (FM) experience cognitive symptoms that negatively impact on their ability to focus, concentrate, organize thoughts, multi-task, or remember information [49]. Objective deficits and perceived cognitive problems (“fibrofog”) are associated with detrimental effects on physical functioning, social participation and quality of life [2,16]. Unsurprisingly then, improving our understanding of cognitive dysfunction in FM and developing effective restorative and/or compensatory treatments is a patient priority [20].

Both presence and intensity of pain have been associated with cognitive dysfunction. Compared to those without chronic pain, poorer cognitive function across multiple domains have been repeatedly identified among those with chronic pain, including attention, processing speed, working memory, long-term memory, and executive function [5,6,7,23,26,36,39,63]. Across chronic pain conditions, including FM, greater pain intensity has been associated with impaired attention and executive functioning, lower mental flexibility, slower processing speed, and poorer visual, verbal, long-term and working memory [21,24,37,42,43,44,56,58]. Conceptual frameworks linking pain and cognitive performance/function, including bottom-up capture of attention by pain and top-down factors (e.g., cognitive goals/task execution), have incorporated gate, load, motivational, and affective theories [32,50]. These overlapping theories have underpinned experiments conducted with healthy adults involving manipulation of phasic and/or tonic painful stimuli. Findings have demonstrated correspondence between increased pain sensation and/or longer pain duration and worse cognitive performance [45,47], and between higher cognitive load and reduced pain perception [17,32,61]. However, given altered brain morphology identified in people with chronic pain [30,35,46], continued comparative examination of the pain-cognitive functioning association in those with and without chronic pain is warranted. This endeavor is supported by a mechanistic model of chronic pain-related cognitive dysfunction that combines theories of competing limited resources and pain-induced changes in neuroplasticity and neurochemistry [19,39,52].

Although, on balance, theory and evidence support the hypothesis that pain and cognitive function are inversely correlated, among those with chronic pain deficits across cognitive functioning domains have not been unequivocally identified [14,34,43,57]. Indeed, positive associations between pain and cognitive function have also been reported [41,55]. Furthermore, a major limitation in this field is that no prior study, to our knowledge, has examined the effect of naturally occurring pain fluctuation on cognitive performance in daily life using a within-person study design. In particular, information is lacking on whether naturally occurring fluctuations in daily pain reach critical thresholds to interfere with cognitive task performance [19,31].

Therefore, we undertook secondary analysis of data from a micro-longitudinal study using ecological momentary assessment (EMA) and objective tests of two specific domains of cognitive function - processing speed (time between task presentation and completion [5]) and working memory (short-term retention and manipulation of information required to complete a task [13]). We aimed to identify whether within-day increase in pain intensity was associated with contemporaneous decrease in cognitive function. Given previous within-day temporal association between pain and cognitive function in a different clinical population [28], we also explored the possibility that an increase in pain precedes a decrease in cognitive function and vice versa. Inclusion of both an FM- and non-FM group allowed us to examine whether any effects were specific to a chronic pain state.

Methods

Participants were recruited out of the University of Michigan from patient registries, health centers, community settings, and using a university-based study recruitment website (www.UMHealthresearch.org). Inclusion criteria were being 18 years of age or older and being fluent in spoken and written English at 6th grade level. Participants with FM had to fulfil 2016 American College of Rheumatology (ACR) FM survey criteria [62]. Exclusion criteria were having a comorbid neurological disorder, learning disorder, or cognitive impairment, current alcohol or recreational drug dependence or prolonged (≥5 years) history of substance dependence, visual or hearing impairment that would impede cognitive testing, having a diagnosis of untreated obstructive sleep apnea, or an atypical sleep/wake pattern (e.g. a shift work schedule). Participants with and without FM were matched on sex, age, and years of education.

Study Procedures

After study procedures were approved by the Medical Institutional Review Board at the University of Michigan, volunteers were screened for eligibility over the telephone. Those who screened as eligible were invited to a face-to-face visit where written informed consent was provided. Data were collected between January and August 2018.

Participants attended a 90-minute visit at the start of the study, followed by a home monitoring period of at least 8 days. At the baseline visit, participants received instruction on how to use study devices (a wrist-worn accelerometer and smart phone) and completed self-report questionnaires and a cognitive testing protocol. At the end of the home monitoring period, participants returned study devices using a postage-paid return box. Participants were compensated up to $175.

The smartphone (ZTE Axon 7 mini; 5.2 inch display; 1080 × 1920 pixels) was programmed with an app that administered cognitive tests and EMA. Participants were instructed to initiate the first of five within-day tests/assessments upon waking. For the following four within-day sessions the phone was programmed to play an alert to prompt the participant. Alerts were programmed on a quasi-random schedule based on each participant’s typical wake time. Scheduled interval prompts ranged between 3–4.5 hours. The data collection and within-day assessment scheduled is outlined in Figure 1.

Figure 1.

Data collection and assessment schedule

Measures

At baseline, participants completed a self-report demographic survey (sex, age, race/ethnicity, number of years of education) and four National Institutes of Health (NIH) Toolbox cognitive tests [22], administered using the NIH Toolbox iPad App [8]. Results have been reported previously [29].

Cognitive tests

Two brief, valid, and reliable cognitive tests [48] were administered using a study-specific smart phone app (fully described in an earlier publication [29]). Participants were not explicitly told that they were doing a ‘processing speed’ or a ‘memory task’ but were told that they would be completing ‘cognitive (thinking) tasks’ each day. In brief, the Symbol Search task assesses processing speed. Participants are shown a row of four symbol pairs at the top of the screen and two symbol pairs at the bottom. Symbol pairs at the bottom have to be matched to a pair at the top as quickly as possible (Figure 2). Sixteen trials were administered per testing session, with reaction times (milliseconds) and errors recorded. Two variables were calculated for trials where session accuracy was ≥70% (a conservative threshold to distinguish poor effort, consistent with procedures applied in a validation study of these measures [48]): mean reaction time per session and reaction time standard deviation (SD) per session. The Dot Memory task assesses working memory. Participants are asked to remember the location of three red dots in a 5X5 grid. After 3 seconds, the grid disappears and the participant is asked to locate and touch letter Fs in an array of Es. An empty 5X5 grid is then presented and the participant is asked to place red dots in the correct locations (Figure 2). Two variables were derived for each trial as a measure of working memory: mean error (mean of the collective Euclidian distance of the three dots from their correct locations), and error SD.

Figure 2.

Objective tests of cognitive function

Ecological momentary assessment

Current pain intensity was assessed on a 0–100 rating scale (0=no pain; 100=worst pain imaginable). Perceived cognitive functioning was assessed with two items from the PROMIS Applied General Concerns item bank [12], adapted for momentary assessment. The items, “How slow is your thinking right now?”, (0–100 rating scale: 0=my thinking is very fast; 100=my thinking is very slow) and “How foggy is your thinking right now?” (0=my thinking is very clear; 100=my thinking is very foggy) were averaged. Higher scores indicated worse perceived cognitive dysfunction (sample Cronbach’s alpha=0.95). At the first within-day EMA, participants were also asked to indicate how refreshed they felt upon waking (0–100 scale, higher scores representing more restorative sleep). As part of the momentary testing schedule participants were also asked to indicate whether they were currently (i.e. at that assessment moment) using alcohol, cannabis, caffeine, narcotics/opioids, nicotine, prescription drugs, or over the counter medications (yes or no).

Data analysis

Demographics, pain intensity and cognitive functioning at baseline were summarized (mean, SD). Day 1 of data collection was discarded as a training day, with the following 7 days of data used for analysis. Five cognitive functioning variables were investigated in separate analyses spanning three domains: self-reported cognitive function, processing speed (mean reaction time; reaction time SD), and working memory (mean error score; error score SD). All analyses focused on within-day associations, with separate three-level multilevel models specified for the 5 outcomes. Assessments were nested within days within participants. To estimate associations between pain intensity and cognitive functioning, person-centered variables were created. These reflected momentary deviations away from the participant’s aggregate mean (centered at 0). Positive values indicate a higher than average score for that participant; negative values indicate a lower than average score for that participant. To identify whether increase in pain intensity was associated with contemporaneous change in cognitive functioning variables, models included fixed person-centered pain-by-FM group interaction terms, and a quadratic fixed effect for time. If the quadratic time function was not statistically significant (p<0.05) it was removed and a linear time effect assumed. This controlled for the effect of time of day on testing (e.g. morning versus evening). To identify whether, within a day, increases in pain intensity preceded increases in cognitive dysfunction, models included fixed person-centered pain scores from the previous within-day assessment-by-FM group interaction (i.e. lag-1). To identify whether increases in cognitive function preceded decreases in pain intensity, models included fixed person-centered cognitive functioning scores from the previous within-day assessment-by-FM group interaction. All models were adjusted for age, sex, years of education, and level of refreshment on waking. Participant and day were included as random effects, and restricted maximum likelihood estimation accounted for missing data under a missing at random assumption. In three separate sensitivity analyses for all models we adjusted for current substance use (total count); current sedative use (alcohol, narcotics/opioids, or cannabis: yes or no), and current stimulant use (caffeine or nicotine: yes or no).

Results

Of 122 people screened for inclusion in the parent study, 16 did not meet eligibility criteria, three declined to participate and three did not complete the data collection protocol. 50 people with FM and 50 people without FM were recruited and completed the study and provided data for this analysis. Mean age was 45.1 (SD 13.9), 88% were female, and the majority were white (81%). Full sample characterization has been published previously [29]. Descriptive aggregate statistics for primary variables are presented in Table 1. Plots of average daily fluctuations in pain intensity and cognitive function by FM status are provided in Supplementary Figures S1 & S2. Of a maximum 3,500 pain intensity assessments across 7 days, 90.2% were complete (FM group 90.3%; non-FM group 90.1%). Similar levels of completion were observed for subjective cognitive function assessments (FM group 90.3%; non-FM group 90.0%) and objective cognitive tests (Symbol search: FM group 88.8%; non-FM group 89.3%. Dot memory: FM group 90.2%; non-FM group 89.9%). Accuracy on tests of cognitive performance was >70% for 98.8% of all tests.

Table 1.

Pain intensity and cognitive function variables: Descriptive statistics (aggregate) by fibromyalgia status

		FM group N=50	Non-FM group N=50
		Mean (SD)
Pain intensity (0–100)		48.7 (17.3)	8.3 (10.3)
Self-reported cognitive dysfunction (0–100)		48.8 (16.7)	17.2 (14.9)
Processing speed Symbol search task	Mean response time (millisecs)	2447.4 (752.6)	2261.1 (618.8)
	SD of response time (millisecs)	1029.3 (347.8)	917.3 (324.9)
Working memory Dot memory task	Error mean (Euclidean distance)	1.6 (0.9)	1.0 (0.7)
	Error SD (Euclidean distance)	1.2 (0.3)	1.0 (0.4)

Associations between within-day fluctuations in pain intensity and cognitive function

Table 2 presents associations between contemporaneous fluctuations in pain intensity and cognitive functioning. Our pre-specified research question was concerned with associations and a possible moderating effect of FM status. Therefore, FM group-by-person-centered pain intensity interaction terms were retained in models regardless of statistical significance. Consequently, regression coefficients for person-centered pain intensity are interpreted as the association of person-centered pain intensity with the cognitive outcome for the reference group. Results in Tables 2, 3 and Supplementary Table S1 are from models that use the FM group as the reference group. Estimates for non-FM group effects presented in the text below were derived after reverse coding.

Table 2.

The within-day momentary association between pain intensity and cognitive function

Fixed effects
	Self-reported cognitive dysfunction	Processing speed: Symbol Search task		Working memory: Dot Memory task
		Mean response time	SD of response time	Error mean	Error SD
Person-centered pain intensity	0.27 (0.22,0.32) p<0.001	1.73 (0.07, 3.39) p=0.04	0.56 (−0.71, 1.83) p=0.39	0.0002 (−0.003, 0.003) p=0.91	0.0008 (−0.002, 0.003) p=0.50
FM group	Reference group
Non-FM group	−30.16 (−36.12,24.19) p<0.001	−189.32 (−420.09, 41.45) p=0.11	−112.00 (−235.25, 11.25) p=0.08	−0.50 (−0.81, −0.19) p=0.001	−0.21 (−0.35, −0.07) p=0.003
Group*person-centered pain intensity	−0.11 (−0.19, −0.03) p=0.01	−4.39 (−7.07, −1.71) p=0.001	−2.55 (−4.60, −0.49) p=0.02	−0.002 (−0.006, 0.002) p=0.37	−0.004 (−0.008, −0.0003) p=0.03
Random effects
Participant	222.09 (33.48)	330989.8 (49927.35)	92728.58 (14214.56)	0.59 (0.09)	0.11 (0.02)
Day	15.29 (3.25)	48596.78 (5224.81)	9403.11 (2110.61)	0.07 (0.01)	0.03 (0.007)
Residual	171.52 (4.88)	168519.2 (4853.18)	109559.9 (3150.48)	0.50 (0.01)	0.38 (0.01)

All values in Fixed effects cells: Regression coefficient, 95%CI, p value; all values in Random effects cells: Estimate, standard error.

Models adjusted for age, sex, number of years of education, perception of feeling refreshed upon awakening (0–100 scale), and a time function (quadratic if significant, p<0.05).

Table 3.

The association between cognitive function and within-day next time point pain intensity

Fixed effects. Dependent variable: EMA Pain intensity; Column headings denote within-day lag 1 time point person-centered cognitive independent variables
	Self-reported cognitive function	Symbol Search (Processing speed)		Dot Memory (Working memory)
		Mean response time	SD of response time	Error mean	Error SD
Lag 1 Person-centered cognitive variable	0.02 (−0.03, 0.07) p=0.36	−0.0003 (−0.002, 0.001) p=0.63	0.0004 (−0.002, 0.002) p=0.71	1.07 (0.18, 1.95) p=0.02	0.78 (−0.28, 1.84) p=0.15
Group*Lag 1 Person-centered cognitive variable	0.02 (−0.05, 0.08) p=0.60	−0.0005 (−0.003, 0.002) p=0.66	−0.001 (−0.004, 0.002) p=0.47	0.23 (−1.10, 1.56) p=0.74	−0.57 (−2.11, 0.97) p=0.47
Person-centered cognitive variable	0.31 (0.26, 0.36) p<0.001	0.002 (0.0003, 0.003) p=0.02	0.0008 (−0.001, 0.003) p=0.42	−0.47 (−1.38, 0.44) p=0.31	0.12 (−0.97, 1.21) p=0.82
FM group	Reference group
Non-FM group	−41.85 (−47.46, −36.25) p<0.001	−41.30 (−46.86, −35.73) p<0.001	−41.20 (−46.76, −35.63) p<0.001	−41.18 (−46.77, −35.59) p<0.001	−41.20 (−46.79, −35.61) P<0.001
Group*person-centered cognitive variable	−0.24 (−0.31, −0.16) p<0.001	−0.003 (−0.005, −0.0009) p=0.006	−0.002 (−0.005, 0.0007) p=0.14	−0.12 (−1.45, 1.21) p=0.86	−1.35 (−2.90, 0.20) p=0.09
Random effects
Participant	192.23 (29.35)	188.23 (28.93)	188.28 (28.93)	190.00 (29.16)	190.13 (29.18)
Day	30.96 (3.96)	34.71 (4.37)	34.53 (4.36)	35.83 (4.37)	35.43 (4.34)
Residual	106.63 (3.74)	112.77 (4.01)	113.26 (4.02)	112.92 (3.96)	113.35 (3.98)

All values in Fixed effects cells: Regression coefficient, 95%CI, p value. All values in Random effects cells: Estimate, standard error.

Models adjusted for age, sex, number of years of education, perception of feeling refreshed upon awakening (0–100 scale), and a time function (quadratic if significant, p<0.05).

For the FM group, each 1-point increase in pain intensity above the participant’s study average was associated with a 0.27 momentary increase in cognitive dysfunction (95%CI 0.22, 0.32). This compared to a 0.17 increase in cognitive dysfunction (95%CI 0.10, 0.23) for the non-FM group. There was a significant group-by-pain interaction, with the FM group exhibiting a steeper slope (Figure 3).

Figure 3.

The moderating effect of fibromyalgia status on the momentary within-day association between pain intensity and self-reported cognitive dysfunction

An increase in pain intensity was associated with longer processing speed for the FM group; each 1 point higher in pain intensity was associated with a 1.73millisec longer mean response time (95%CI 0.07, 3.39). In contrast, for the non-FM group, higher pain intensity was associated with a shorter mean response time (−2.66millisecs, 95%CI −4.77, −0.56). A significant group-by-pain interaction term supported a case for moderation (Table 2, Figure 4A). Although pain intensity was not associated with increased variability in response times for those with FM, it was associated with significant reduction in variability for the non-FM group (B=−1.99, 95%CI −3.61, −0.37). A significant interaction term supported a case for moderation (Figure 4B).

Figure 4.

The moderating effect of fibromyalgia status on the momentary within-day association between pain intensity and objective cognitive function

Fluctuations in pain intensity were not significantly associated with working memory (mean errors or their SD) for the FM group. However, there was a significant FM group-by-pain interaction for the variability outcome that could be explained by the non-FM group exhibiting lower variability in the presence of higher pain (B=−0.003, 95%CI −0.006, −0.0003) (Figure 4C).

Do increases in pain intensity precede increases in cognitive dysfunction?

Results of analysis of within-day previous time point pain intensity on current cognitive function are presented in Supplementary Table S1. Models include a lag-1 person-centered pain variable and a FM group-by-lag 1 person-centered pain intensity interaction term, in addition to the pain and FM group-by-pain interaction term from the previous ‘contemporaneous’ model. This facilitated assessment of person-centered pain-intensity on next time point cognitive function over and above momentary associations. There was no evidence that increases in pain temporally preceded change in subjective or objective cognitive function for the FM group. For the non-FM group, a 1-point increase in pain intensity was associated with a 0.12 increase in self-reported cognitive dysfunction at the next within-day assessment (B=0.12, 95%CI 0.04, 0.20), and a reduction in working memory variability (B=−0.005, 95%CI −0.0009, −0.001). In both cases there was no evidence for moderation by FM status.

Do increases in cognitive function precede decreases in pain intensity?

Table 3 presents associations between previous within-day cognitive function assessment and current pain. For the FM group, only one variable was associated with within-day next assessment pain intensity; mean error on the Dot Memory task (working memory). Each one-point increase in person-centered error was associated with a 1.07 (95%CI 0.18, 1.95) increase in pain intensity 3–4.5 hours later. There was no evidence of a moderating effect of FM. For the non-FM group, significant associations were observed between an increase in self-reported cognitive dysfunction and within-day next pain assessment (B=0.07, 95%CI 0.02, 0.13), and each 1-point increase in person-centered variability (person-centered error SD) on the Dot memory task (working memory) was associated with a 1.23 reduction in within-day next assessment of pain intensity (95%CI −2.33, −0.13). In both cases there was no evidence for a moderating effect of FM status.

All results reported above were robust to separate sensitivity analyses that also adjusted for i) total substance use count, ii) sedative use, and iii) stimulant use, with negligible impact on effect estimates and no change to statistical significance or interpretation.

Discussion

We aimed to ascertain whether within-day increases in pain intensity are associated with contemporaneous increases in objective and subjective cognitive dysfunction among people living with FM, and whether increases in pain intensity precede increases in cognitive dysfunction or vice versa. We found that momentary increases in pain intensity are accompanied by increases in self-reported cognitive dysfunction and by a reduction in objectively measured processing speed, but not working memory. This builds on previous cross-sectional research that has demonstrated chronic pain-related deficits in processing speed and selective attention but not attention switching or short-term working memory [14,18].

When compared to a non-FM group, although the direction of the momentary association between pain and self-reported cognitive function was the same, effect magnitude was augmented among those with FM. In contrast, for tests of processing speed and working memory, direction of effects differed. Moments of higher pain associated with faster and less variable processing speed, and less variability in working memory for the non-FM group. We trimmed our dataset to ensure that ≥70% of trials were correctly completed, therefore this cannot be attributed to those without FM quickly/impulsively completing tasks without concern for accuracy. However, while correct completion arguably reflects task engagement, future study would benefit from more complex tasks that challenge sustained engagement and affective-motivational dynamics pertinent to those with FM [54]. Although between-group differences may be statistical artefacts given lower variability and more instances of no pain in the non-FM group, our results tentatively suggest that in a non-FM context, acute pain may sharpen focus. In the absence of evidence from other ambulatory studies, this interpretation may be compared to findings from experiments conducted with healthy adults. Induced pain unrelated to a cognitive task has been observed to only be prioritized when it is perceived as threatening [19,53], and faster completion of a test of attentional control and executive function have been recorded in the presence of an electrocutaneous pain stimulus [3]. It has been argued the latter finding may be attributable to the pain stimulus acting as a cue to task completion [38]. However, this does not apply in our study. Conducted in a real-world context, our results support a case for a positive pain intensity-cognitive function association in those without chronic pain. It is possible that the moderating effect of FM status on pain-cognitive function associations that we have identified may reflect a dampening among those with fibromyalgia of an otherwise adaptive physiological response [11].

Our observation of momentary associations between pain intensity and subjectively reported and objectively cognitive function among those with FM may be an expression of intertwined nociceptive and cognitive mechanisms, including chronic pain-induced alterations in neuromediators and neural networks [25], and common neuroanatomical regions involved in pain processing and cognitive functioning (e.g. the anterior cingulate cortex [10,27], insular cortex [4,9,33] and periaqueductal gray [51]). An argument for synchronized effects in our study is supported by lack of strong evidence for a temporal relationship between the two constructs. However, a temporal effect cannot be ruled out as the average duration of 3.5–4 hours between within-day assessments may be too long for effects to be sustained. Collecting data more intensively may alleviate this. However, participant burden and the intrusiveness of such highly regular testing likely make this infeasible. This is difficult to resolve and may require a novel study design, for example scheduling EMA/cognitive testing more closely at specific points during the day (e.g. one hour apart, three times a day). As it stands, our data indicate that pain and cognitive dysfunction covary simultaneously. Admittedly, in absolute terms, the effects sizes of the momentary associations were small. While interpretation of results garnered from micro-longitudinal data collection require consideration of the extent to which even “miniscule” moment-to-moment effects may accumulate to result in larger effects over time [1,15], we were interested in momentary effects. Accordingly, our finding that a reduction in working memory precedes increase in later pain intensity among people with FM should be interpreted cautiously. However, we believe it to be deserving of further examination as it replicates a pattern identified in a within-person study of perceived cognitive dysfunction and pain intensity conducted with people with Multiple Sclerosis [28], and may hold prognostic and/or therapeutic potential. It is possible that deficits in working memory may be an early sign that a pain flare is imminent and compensatory strategies need to be initiated. It is also possible that interventions that optimize working memory may diminish within-day pain intensity variability.

Study strengths include use of reliable and valid measures of objective cognitive function alongside a subjective measure, high levels of complete data, and inclusion of a non-FM control group. Although use of ambulatory cognitive assessment conferred ecological validity, assessment integrity is a potential limitation; although participants were instructed not to use external aids (e.g., pencil and paper), we have no way of knowing if this instruction was followed. Regarding other study limitations, although we were able to examine two objective measures of cognitive functioning (processing speed and working memory), our tests were brief and based within the visuospatial domain. Understanding would be enhanced by investigating other domains associated with chronic pain, e.g. executive function [56], and verbally mediated test performance may show different effects. We assessed the influence of within-day one time point lag between pain and cognitive functioning. We could have examined longer within-day durations (i.e. lag-2, lag-3), however, given lack of evidence for a lag-1 effect, this was considered unnecessary. Also, as we were interested in within-day effects, we did not examine previous day’s pain on next-day cognitive functioning or vice versa. It is notable, though, that on partitioning variability in key variables, very little variance was attributable to between-day differences, regardless of FM status.

We decided, a priori, not to adjust for fatigue or symptoms of depression or anxiety and understand that this may be perceived as a study weakness. We align with a position that conceive of these as integral to FM multi-symptomology [39]. By controlling for these factors, we would have removed ‘fibromyalgianess’ and believe this would have been counterproductive. From a statistical perspective, these symptoms are just as likely to be mediators of the pain-cognition association as they are confounders. Adjustment would have eradicated any indirect effects transmitted through these variables. We did, however, adjust for level of refreshment on awakening, given established associations with pain intensity [59], and cognitive functioning [60]. As perception of sleep restorativeness preceded within-day associations, it could only have been a confounder. Also, although we adjusted for substance use in sensitivity analyses, pharmacological effects cannot be ruled out as data were not available regarding intake of non-opioid analgesics known to affect cognitive function (e.g., pregabalin) [64]. Age was adjusted for in all analyses, however, given previous identification of a moderating effect of age on the pain-cognitive function association [40], in post-hoc, unplanned analysis, we examined this in the FM group. Moments of higher pain were more strongly associated with greater perceived cognitive dysfunction (but not objective test performance) with increasing age. Further research into the impact of age, and any age-specific differential effect on subjective and objective measures of cognitive function is therefore warranted.

The sole criteria differentiating the FM group from the non-FM group was meeting the 2016 ACR FM survey criteria. Information about pain history prior to the study or source of pain during the study was not collected. As such, our comparison is the association between those with versus without a centralized (widespread) pain condition. As the first (to our knowledge) study of ambulatory cognition in FM, the control group was included to provide context for what ‘normative’ cognitive functioning in the context of ambulatory cognitive tests. However, more thorough characterization of the pain history and experiences of any non-FM control group are recommended in future study. Furthermore, comparison of the pain intensity-cognitive function association between people living with FM and people living with other pain conditions with more ‘bottom-up’ features (e.g. chronic low back pain, osteoarthritis) could illuminate differences that may be ascribed to underlying pain mechanisms.

In summary, we have demonstrated that within-day momentary increases in pain intensity are associated with increased self-reported cognitive dysfunction in people with and without FM, and that the association is significantly stronger in FM. We have also shown that momentary increases in pain intensity are associated with a reduction in processing speed in FM, but an increase in processing speed in a non-FM group. Although pain intensity was not associated with momentary processing speed variability or working memory variability in FM, it was associated with less variability in both cognitive constructs in a non-FM group. The only evidence to support an argument that increases in pain intensity precede within-day change in cognitive function came from those without FM, where increases in pain intensity preceded increases in self-reported cognitive dysfunction, and a reduction in working memory variability. Evidence in favor of cognitive dysfunction preceding pain intensity exacerbation was derived from an FM group, where reduction in working memory preceded within-day increase in pain intensity, and from a non-FM group for both self-reported cognitive dysfunction and increased working memory variability preceding an increase in pain intensity.