The predictive role of pain catastrophising following genicular arterial embolisation for the treatment of mild and moderate knee osteoarthritis

Introduction

It is estimated across a lifespan that 47% of women and 40% of men will develop symptomatic knee osteoarthritis[39]. Osteoarthritis (OA) is the most common form of arthritis[31], which in turn is a leading cause of disability globally[22]. At present, OA is not considered to be a curable disease, in part, because the pathophysiological mechanism is not yet comprehensively understood[36]. The goal of OA treatment is to slow the progression and alleviate the symptoms of the disease.

Traditionally, initial treatment consists of conservative options such as physiotherapy, orthotics and pharmacology. It is only when all conversative options are exhausted, and the pathological indicators reach a threshold of severity, that surgical options are considered[13,38]. Given the central role of psychosocial factors in the maintenance of chronic knee pain[37], a prolonged period of poor treatment response can have serious implications for the entrenchment of chronic pain. Mild-to-moderate OA resistant to nonsurgical options, yet not severe enough to warrant join replacement surgery, poses a significant management problem. Genicular artery embolisation (GAE) is a novel interventional radiological technique that is easier to deliver and less invasive which is designed to subvert neoangiogenesis within the joint, hypothesised to contribute to structural damage and pain in knee OA [30,34,40].

While promising, preliminary data for GAE indicates a subset of patients who do not respond to treatment, despite a technically successful procedure [34]. These findings are also mirrored within more invasive surgical options such as total knee replacement (TKR), wherein 6-30% of patients continue to experience chronic pain after the procedure[2,7,10]. It is clear that a strictly pathophysiological approach to the assessment of pain is ineffective[8] and that to understand the interpatient variability in outcomes we must complement these assessments with psychosocial insight. Moreover, complementing these assessments with predictive perspectives may facilitate earlier intervention, to help disrupt the negative progression of the chronic pain cycle.

Pain catastrophising (PC) is a cognitive-affective bias characterised by a negative interpretation of the consequences of pain[42,52]. Catastrophising is regularly quantified using the Pain Catastrophising Scale[51], and comprises of three main elements; rumination, magnification and helplessness. PC has been associated with lower pain thresholds[16] and poor longitudinal surgical response to a range of conditions[28,41], including knee osteoarthritis[10,15,43]. Alongside it’s predictive capabilities, it has been shown to be modifiable by psychological intervention[17,18,33], underlining the potential value for presurgical identification to improve outcomes.

PC is frequently characterised as a cognitive attentional bias to the processing of nociceptive stimuli. Evidence suggests catastrophising is not directly associated with the sensory-discriminative dimensions of pain, but instead the modulation and processing of the unpleasantness of pain[46]. Attentional regulation has been found to also play a key role in catastrophising, with high catastrophisers often struggling to disengage attention from pain[12]. As such, neuroscientific investigation focuses on regions such as the dorsolateral prefrontal (DLPFC), anterior cingulate (ACC), insula and medial prefrontal cortices. These cortical regions are associated with the emotional, and attentional modulation of pain, as well as pain salience, vigilance and awareness. Specifically, the dlPFC is thought to play a role in top-down modulation, underlying the facilitatory influence of PC on pain[19,24,35,47,48]. Therefore, while PC is not directly associated with the sensory-discriminative response to pain, the processing and suppression of this sensory information may underlie the maladaptive influence of catastrophising[42,46]. Furthermore, the manner in which individuals attend to pain involves an interaction between top-down and bottom-up influences and is likely to facilitate catastrophisation[12].

In the current study, we examined pre-surgical characteristics of patients and tested the hypothesis that clinical outcomes following GAE in patients with mild to moderate knee OA can be predicted by presurgical pain catastrophising levels. Secondly, we investigated the influence of pain catastrophising on Intrinsic Attention to Pain (IAP), specifically, whether catastrophisers have a higher tendency to attend towards pain. Lastly, to better understand intrinsic neural mechanisms, we used resting-state fMRI (rs-fMRI) to examine dlPFC-to-whole brain neural connectivity at rest. We hypothesised that the neural mechanism underlying PC would be associated with variable connectivity of the dlPFC, as a key region in pain modulation and the suppression of pain intensity[35], and regions of the brain associated with sensory processing of nociceptive signals, such as the motor and somatosensory cortices[54,58].

Methods

Materials

fMRI Analysis

ROI Selection

For the purposes of preparing the ROI, a DLPFC mask was identified using the Harvard-Oxford cortical 100% probabilistic structural atlases. Due to no specific hypotheses regarding lateralised neural mechanisms within resting-state, a bilateral mask was then created using this method (Fig.1). The DLPFC was selected as a seed due to its prominent role in top-down modulation of pain[46,47], as well as its association with pain catastrophising[19,48].

• Download figure
• Open in new tab

Figure 1

Bilateral dorsolateral prefrontal cortex seed across axial (top) and sagittal (bottom) planes, with co-ordinates shown in mm

• Download figure
• Open in new tab

Figure 2 Longitudinal outcomes following successful completion of genicular arterial embolisation (standard error bars).

| Lower WOMAC and NRS values represent less severe impact of OA and lower pain, respectively. Stars denote significant improvements from baseline

Pre-processing

All analysis was performed using FMRIB’s Software Library Package (FSL 6.0[27]), following the Component Based Noise Correction Method (CompCor[4]). During acquisition, the first five volumes were discarded to facilitate signal equilibration. Correction for interleaved acquisition was applied and data were spatially smoothed using a 5mm full-width half-maximum (FWHM) gaussian kernel. The Brain Extraction Tool (BET[49]) was used for skull stripping. MCFLIRT[26] was used for the purposes of motion correction, and data were visually inspected to identify problems with registration, inadequate skull-stripping or uncorrected motion artifacts. To reduce the influence of non-neuronal activity, FAST[59] was then used to created segmented white matter (WM) and cerebrospinal fluid (CSF) masks, which were then thresholded to 0.99. Time-series were then extracted for each participant and added to the GLM as nuisance variables. Residuals from pre-processing and nuisance removal were then normalised and band-pass filtered (0.1/0.001Hz) to reduce the influence of high-frequency (i.e. cardiac/respiratory) and low-frequency (i.e. scanner drift) factors.

Resting-state Connectivity Analysis

The standardised DLPFC mask was registered to single-subject space, and a mean time series of all voxels within the ROI were extracted and added as an explanatory variable within a whole-brain functional connectivity analysis. Resulting contrast maps were then used as inputs within a higher-level analysis, alongside patient’s normalised pain catastrophising scores. The purpose of this analysis was to identify regions were connectivity to the DLPFC was associated with individual differences in PCS. Using FEATQuery, parameter estimates were extracted from resulting significant clusters for the purposes of the graphical representation of connectivity. Multiple comparisons corrections were applied using the Gaussian random field theory (Z<2.3;p<.05).

Results

Presurgical Baseline Psychometrics

Pain catastrophising scores were significantly correlated with all other baseline measures, with the exception of NRS. At baseline, pain catastrophising was associated with poor sleep quality (r(29)=.40,p=.03), low trait mindfulness (r(28)= −.45,p=.02), depression (r(29)=.41,p=.03), anxiety (r(29)= .53,p=.003) and higher pain via the KOOS (r(30)= −.45,p=.01) and WOMAC (r(30)=.39,p=.03), whereas the NRS was a non-significant trend (r(30)=.29,p=.12). PCS was also significantly correlated with IAP (r(30)=.53,p=.002), indicating that higher catastrophisers are more likely to attend to pain (Fig.3).

• Download figure
• Open in new tab

Figure 3 Association between pain catastrophising and intrinsic attention to pain

| Positive values on IAP represent attention to pain, negative values represent attention to something else.

View this table:

• View inline
• View popup
• Download powerpoint

Table 1 Association between presurgical baseline variables

Pain Catastrophising and postsurgical outcomes

Pain catastrophising significantly predicted reductions in pain at 6-weeks (F(1,27)=4.53,p=.043,R²=.14,R² =.11, SE_β= .41), 3-months (F(1,28)=12.1, p<.005, R²=.30,R² =.28, SE_β= .35), 12-months (F(1,25)=4.61,p=.042, R²=.16,R² =.12, SE_β= .43), and 24-months (F(1,17)=12.01,p<.005, R²=.41,R² =.38, SE_β= .46). This suggests that patients who are high catastrophisers stand to gain the most beneficial reductions in pain following embolisation (Fig 4.).

• Download figure
• Open in new tab

Figure 4

The association between baseline pain catastrophising, and reductions in baseline pain at 6-weeks, 3-months, 12-months and 24-months(clockwise). | Lower values represent decrease in WOMAC score over time, and improvement regarding OA impact)

Association of Pain Catastrophising and functional connectivity of the dorsolateral frontal cortex

Analysis of rs-fMRI data revealed that patients with higher PCS scores were associated with higher connectivity between the dorsolateral prefrontal cortices and the somatosensory, motor, premotor, insula, operculum, anterior cingulate cortices (Fig.5; table 2).

• Download figure
• Open in new tab

Figure 5

Clusters associated with connectivity to the DLPFC and pain catastrophising in MNI space (z<2.3) | Figure numbers relate to co-ordinates described in Table 2.

View this table:

• View inline
• View popup
• Download powerpoint

Table 2 Statistical peaks of clusters in MNI space associated with pain catastrophising and functional connectivity to the DLPFC

Discussion

The study investigated if pain catastrophising can be used to predict clinical outcomes following genicular embolisation for the treatment of mild-moderate osteoarthritis. We also investigated how individual differences in catastrophising were associated with variations in functional connectivity of the DLPFC, a key pain modulatory region frequently associated with the underlying mechanisms of pain catastrophising[19,35,46,48]. We observed that, on average, patients experienced lasting reductions in pain as a result of the procedure, but that unexpectedly, those who were high catastrophisers at baseline gained the most profound improvements at all time-points (6-weeks, 3-months, 12-months and 24-months). Further, individual differences in PCS scores at baseline were associated with worse pain and a myriad of negative psychological impacts at baseline, as well as variations in functional connectivity at rest. We showed that pain catastrophising was associated with higher connectivity between the DLPFC and the anterior cingulate, premotor, motor, insula, operculum and somatosensory cortices.

Pain catastrophising is described as a set of maladaptive cognitions characterised by heightened pain intensity and unpleasantness[52,53], as well as an inability to disengage from the experience of pain[12]. Interestingly, our data indicate that high catastrophisers gained the most substantial reductions in pain following embolisation. For catastrophisers, pain represents an irrepressible aversive influence which cannot be disengaged from, as supported by our finding that PCS correlates with IAP scores. At baseline, catastrophising was associated with higher osteoarthritis pain, as well as associative negative lifestyle and psychological influences, such as depression, anxiety and poor sleep. The significant postsurgical decreases in pain may be most beneficial for those who find the impact of pain to be more intrusive, intense and unpleasant.

Interestingly, catastrophising has previously been linked to poor long-term outcomes following surgical intervention for osteoarthritic pain[21,43]. An important distinction between this experiment and similar previous studies, is that this study comprised solely of patients with mild-moderate osteoarthritis. Catastrophising has previously been shown to predict poor outcomes to invasive surgical procedures for knee osteoarthritis such as arthroplasty, often reserved for older patients with more severe or debilitating osteoarthritis [13], who are especially vulnerable to catastrophising[60]. Despite this, multiple studies have reported no association and challenged this position, stating the PC may be less trait-like and robust than initially thought[9,25]. One study investigating PC and recovery following total (TKA) and unicompartmental (UKA) joint replacement in 615 patients reported a similar finding to this study that high catastrophisers were associated with greater improvements on the Oxford Knee Score (OKS) scale[6]. Despite the lack of clarity on the direct influence of PC on postsurgical OA pain, it is logical that patients who have suffered for a longer duration from more intense and frequent pain may be at risk of more robust and entrenched pain and negative affect, whereas earlier treatment of mild-moderate OA stands to have more chance of malleable and impactful benefit.

Previous studies have also identified a moderating effect of treatment efficacy on the relationship between pain catastrophising and post-surgical pain[56]. Therefore, it may be that the beneficial outcomes for catastrophisers in this sample are because the successful treatment of their knee had a bifold impact on reductions in pain and pain catastrophising, as well. Pain catastrophising is often described as a robust, cognitive bias, representing a stable individual difference[52], although this position has been challenged more recently. It has been proposed that catastrophising may be a dynamic construct related to pain intensity[55], supported by its high malleability across varying interventions for surgical patients[18]. Successful clinical treatment may theoretically alter responses to multiple items in the PCS such as “It’s terrible and I think it’s never going to get any better” (item 3) or “there’s nothing I can do to reduce the intensity of pain” (item 12)[51]. In these instances, the entrenchment of pain may not yet have taken hold, and maladaptive cognitive biases can still be challenged via successful alleviation of pain. However, as this study only collected PCS scores at baseline, future studies should investigate the impact of successful clinical intervention on the stability of PCS scores for patients with non-severe OA.

Alongside the behavioural findings of PCS, this study also provides insight to the neural mechanisms underlying individual differences in catastrophising. Patients completed rs-fMRI scans, where no task was administered, to examine intrinsic functional connectivity. Evidence has frequently identified the DLPFC as a key region in the process of catastrophising, associated with the interpretation of pain and descending pain modulation[46–48]. Our data indicates that catastrophising is associated with higher connectivity between the DLPFC, and multiple areas of the brain associated with sensory-discriminative processing of pain (motor, sensorimotor, premotor cortices), attentive-perception and salience of pain (insula and anterior cingulate cortices) and modulation of pain encoded within the brainstem (medulla and PAG).

The interpretation of rs-fMRI findings must be evaluated conservatively, as many direct explanations of pain processing would require event-related stimulation to empirically test. Additionally, this finding requires replication due to the relatively low sample size[20], and this follow-up will be completed in the second phase of the study. However, as catastrophising facilitates heightened salience and attentional focus towards pain, increased functional connectivity of the DLPFC with regions involved in processing of pain may represent an increased demand for endogenous modulation. These findings complement previous work by Seminowicz and Davis[46], who found that during mild painful stimulation, PCS was associated with increases in activity in the DLPFC, as well as in the insula, motor and rostral anterior cingulate cortices, matching clusters identified from our analysis. Conversely, if the intensity of pain is increased, PCS was then associated with decreased activity in the DLPFC. The authors proposed that during intense pain, high catastrophisers may have difficulty disengaging from pain, via a lack of top-down control.

This dynamic relationship between PCS and pain intensity may provide an explanation as to why PCS is associated with poor outcomes in severe OA, whereas when mild-moderate OA is treated successfully via embolisation, these patients experience a reduction in pain which they can then continue to effectively modulate.

The current study is concordant with the interim analysis by Little et al[34]., suggesting that genicular arterial embolisation is an effective treatment for mild-moderate osteoarthritis pain. Pre-surgically, pain catastrophising is associated with a range of negative co-morbidities, such as depression, anxiety and increased pain. However, our data suggests that patients who are high catastrophisers stand to gain the most benefit from successful intervention, and the PCS can predict outcomes up to 12-months post-surgically. Neural data suggests that catastrophisers have a higher functional integration of descending modulatory and pain processing regions in the brain, and that the DLPFC is a key region in this process. These results suggest that the increased attentional preoccupation with pain at baseline, facilitates a heightened requirement for pain modulation, which is supported by previous studies[46]. A follow-up trial will aim to replicate these findings within a larger sample in a randomly controlled trial (RCT), alongside sham surgery, to evaluate embolisation alongside a suitable control.