Intact tactile anisotropy despite altered hand perception in complex regional pain syndrome: rethinking the role of the primary sensory cortex in tactile and perceptual dysfunction

Participants

The convenience sample included people with Complex Regional Pain Syndrome (CRPS), controls with pain of other origin (PC), and controls without pain (healthy subjects). This study was approved by the University of South Australia (UniSA) Human Research Ethics Committee (No. 35944). All recruited participants provided written informed consent and the study was performed in accordance with the ethical standards of the Declaration of Helsinki (1991).

Recruitment of all groups occurred at three sites: the School of Health Sciences, UniSA; Royal Melbourne Hospital Rehabilitation Hand Therapy Clinic; Neuroscience Research Australia, Sydney. Patients with pain were recruited using posters displayed at UniSA and in clinical settings, through recruitment appeals on social media or within existing research databases, and via word of mouth. Healthy participants were recruited via institutional students and members of the staff, through public advertisement and included pain participants’ relatives, if interested.

Participants were screened via email or telephone and prior to participation, clinically screened by a trained health professional on the day of testing to assess symptoms/signs and function. Patients with CRPS type I were required to have received an original diagnosis of the condition by a pain specialist and to meet the CRPS research criteria on the day of testing (see Table 2 in results section for frequency of sensory, sudomotor, vasomotor, trophic, and motor signs). PC participants were required to have pain in the upper limb for at least 6 months. Healthy participants were required to be pain-free in the upper limb and have no history of trauma, or medication intake (aside of oral contraceptives). Potential participants, who reported a current diagnosis of a neurological or psychiatric condition, were excluded.

An a priori sample size was calculated using G*power software (Version 3.1). We initially powered this study to detect a small to medium effect (f = 0.2) using a 3 (Group) × 2 (affected versus unaffected limb) repeated measures analysis of variance (RM ANOVA), with 80% power, alpha of 0.05, correlation for repeated measures of 0.5, and adjusting for a 10% withdrawal rate. This resulted in a total of 45 participants (15 participants/group). Given that tactile distance judgements have never before been evaluated in people with CRPS, we were unable to base this effect size directly on past studies. However, we purposefully chose a small to medium effect size because past work evaluating TPD threshold in people with CRPS suggested that powering for such effects was reasonable. Specifically, differences in TPD threshold between CRPS and healthy volunteers were of a moderate to large effect (Cohen’s d = 0.63) and TPD threshold differences between the unaffected and affected limb in CRPS were of a moderate effect (Cohen’s d = 0.53) (Lewis & Schweinhardt, 2012) Our choice was also informed by guidelines for meaningful effect sizes in highly heterogenic clinical populations (Gignac & Szodorai, 2016; Lipsey et al., 2012). Following consultation with a biostatistician, generalised estimating equations (GEE) were chosen to allow a more comprehensive analysis of the data (e.g., to allow consideration of other clinical factors that might influence tactile distance judgements). While the ability to undertake formal power analyses based on GEE is limited, importantly, GEE has been shown to have higher power with lower sample sizes and/or lower numbers of repeated measures than RM ANOVAs (Ma, Mazumdar & Memtsoudis, 2012), suggesting we would be adequately powered.

Questionnaires and assessments

Tactile distance judgement

We used a recently developed anisotropy paradigm that involves judging tactile distances applied across and along the hand (Longo & Haggard, 2011), the area that is frequently reported as the centre area of perceptual distortion in CRPS patients (feels swollen or larger than actual physical size (Lotze & Moseley, 2007; Moseley, 2005; Peltz et al., 2011).

We used a bespoke device (Fig. 1): two wooden posts were mounted on a wooden disk (to assure stable stimulation) at three different distances: two, three or four centimetres (cm) apart. Here, we used three distance combinations (2/3; 2/4; 3/3), evaluating three length differences (0 cm/1 cm/2 cm). For testing this resulted in five distance combinations, based according on which stimulus-pair was provided across versus along the hand axis: (i) 2 across/3 along; (ii) 3 across/2 along; (iii) 2 across /4 along; (iv) 4 across/2 along; (v) 3/3—across/along. In the context of identical tactile distances (3 cm/3 cm), a stimulus presented across the hand dorsum typically feels wider (subsequently described as farther apart) than a stimulus with identically distanced wooden posts but presented along the hand dorsum. This constitutes a perceptual bias resulting from the spatial processing of tactile input and is proposed to mirror the morphology of S1 hand representation (Knight Fle, Longo & Bremner, 2014; Longo, Ghosh & Yahya, 2015; Miller, Longo & Saygin, 2016).

We followed the protocol presented by Longo and Haggard (Longo & Haggard, 2011) that has been used consistently in the literature (Knight Fle, Longo & Bremner, 2014; Longo, Ghosh & Yahya, 2015; Miller, Longo & Saygin, 2016). This validated protocol for tactile distance judgement testing (Longo & Haggard, 2011) involved participants sitting with eyes closed, arms and palms resting flat on a table, fingers splayed. The bespoke device with two wooden posts of varying distance combinations (2, 3, and 4 cm apart) was used to provide tactile stimuli to the hand (termed a stimulus pair). Two different tactile distance stimulus pairs were provided to the dorsum of the hand: one stimuli pair across the hand (perpendicular to the longitudinal axis of the limb) and one stimuli pair along the hand (parallel to the longitudinal axis of the limb). Participants were asked to judge which stimulus pair (of the two provided) felt farther apart. For example, if one stimulus pair consisting of rods a distance of 2 cm apart was delivered across the hand and one stimulus pair with a distance of 3 cm between the two rods was delivered along the hand, participants underwent a forced choice paradigm, identifying which of the two stimuli pairs felt farther apart. Each stimulus pair application lasted for approximately one second with an interstimulus interval (ISI) of approximately one second.

Participants underwent one test session which consisted of two blocks (one block for each hand). During each block, each of the above outlined five distance combinations were presented on the back of the hand, either across the axis of the hand or along the axis of the hand. For each hand, 50 tactile distance judgements were made (5 distance combinations x 10 repetitions). The order of trials was pseudorandomized, and the tested hand order was counterbalanced. This was chosen to minimize attention trade-off and to reduce the risk of increased pain or pressure sensitivity in people with CRPS or hand pain.

Tactile distance judgement outcomes

Two primary outcomes were evaluated to comprehensively investigate processing of spatial features of tactile input: anisotropic perception bias and tactile anisotropy (See Fig. 2).

Data analyses (using IBM SPSS Statistics 24.0)

Study sample

Fifty-three people participated. Upon clinical screening on the day of testing, three CRPS patients were excluded as they did not meet the CRPS research criteria. Two patients with CRPS reported increased pain levels during tactile stimulation such that testing was discontinued on the affected hand. Therefore, complete datasets of both hands were available for 48 participants (CRPS: N = 14; PC: N = 15; healthy subjects: N = 19).

All participants were right-handed, except for one female patient with CRPS and one female patient with rheumatoid arthritis. Both patient groups were predominantly comprised of females (CRPS12/14; PC: 13/15; healthy participants: 9/19). Healthy participants were younger than the patient groups (mean, sd and range: CRPS = 47.6 ± 12.6; 27–56; PC = 58.7 ± 17.1; 27–85; healthy subjects = 36.7 ± 14.5; 17–66; ANOVA main effect F (2,49) = 9.5; p = 0.001). Of patients with PC, one had carpal tunnel syndrome, two reported chronic pain following soft tissue injury, 10 had rheumatoid arthritis, and 2 had osteoarthritis. Table 2 outlines mean ratings and frequencies of sensory, autonomic and motor signs and symptoms of both patient groups.

Pain intensity and duration

Current and average pain intensity reports and duration of pain did not differ between the two patient groups (Table 2). CRPS patients reported significantly higher physical impairment (DASH) than PC patients (ANOVA main effect F(1, 30) = 14.8; p = 0.001).

Body perception disturbance

The feeling of limb foreignness (FLF) differed between CRPS patients and PC patients and was higher for CRPS patients than for those with PC, expressed in a higher Neglect-like score (ANOVA main effect F(1, 30) = 9.1, p = 0.005). Similarly, CRPS patients reported higher distortion of hand perception than PC participants, expressed in a higher score on the Bath Body perception Scale score (ANOVA main effect F(1, 29) = 6.8; p = 0.014).

Hand perception (visual scale task)

Table 3 lists the frequency of chosen hand images that participants identified as matching the perceived size of the own affected (dominant) or non-affected (non-dominant) hand. The chosen template corresponding to the affected hand for patients with CRPS, PC and pain-free participants differed significantly between the three groups (χ2(8) = 16.04, p = 0.042). A post-hoc analysis of the contingency table with Bonferroni corrected p-values showed that pain–free participants were less likely than the pain groups (CRPS and non-CRPS pain) to match their dominant hand perception to a hand template that depicted an increase of 30% (χ2(8) = 8.98, p = 0.002). No group differences were found regarding the non-affected/non-dominant hand (χ2(8) = 7.46, p = 0.87).

Tactile distance judgements

Table 4 reports all proportions and percentages of across-oriented stimulus pairs perceived as farther apart on the skin. Results are structured to first report the overall group comparison and then list predictors that are specific for the patient- group comparison. Further details on the general predictor variable set and odds for each predicting variable for each outcome measure are listed in the supplementary section (Tables 1–3).

Primary analysis: Anisotropic perception bias

Descriptive analyses of response frequency support a clear anisotropic perception bias on dorsal surfaces of both hands in all participants (see Table 4). That is, of all equal distance stimulus pairs (3/3 cm), 69.8% (335/480; 95% CI [65.7–73.9]%) were perceived as farther apart when delivered in the across-orientation on the dorsum of the affected/dominant hand. Similarly, on the non-affected/non-dominant hand, 64.6% (310/480; 95% CI [60.3–68.9]%) of all equal distance stimulus pairs (3/3 cm) were perceived as farther apart when delivered in the across-orientation. In all groups, occurrence of anisotropic perception bias significantly exceeds chance on both hands, (50%, for OR and CI see general set of predictors and supplementary section). The magnitude of anisotropic perception bias was not different between participants with CRPS, PC, and healthy participants (see Table 4 and Table S1).

General set of predictors for all participants

Anisotropic perception bias did not differ between groups (CRPS = OR: 0.501; CI 95% [0.215–1.171], p = 0.111; PC = OR: 0.531; CI 95% [0.244–1.159], p = 0.112; healthy subjects = reference group) or hands (OR: 0.783, CI 95% [0.518–1.182], p = 0.244) when controlling for gender, pain intensity and age (see Table 1 of the supplementary section). In this multivariate analysis, the best model of fit (QUICC: 1193.54) to explain the variance in perceiving across-oriented stimulus pairs as longer in the context of equal stimulus pair lengths (3/3) only included gender, current pain intensity and age. And of these variables, only gender was significant, finding that male participants had 0.5 times the odds of having anisotropic perception bias than female participants (OR: 0.507; CI 95% [0.300–0.857], p = 0.011).

Patient specific clinical variable set

Considering only the pain groups, results show that a combination of gender, presence of allodynia, and current pain intensity best predicts the anisotropic perception bias in the context of equal distance (3/3 cm) stimulus pairs (QUICC 727.009) on a multivariable level. As above, male gender significantly reduced the log odds for anisotropic perception bias (OR 0.609, CI 95% [0.376–0.986], p = 0.044). Current pain intensity was a potential predictor (mean [95% CI] bias = 1.012 [1.000–1.023], p = 0.054). Notably, hand perception did not predict anisotropic perception bias—it was not included in the multivariate model (did not add to model fit), and univariate analyses were not statistically significant (Body perception scale: OR: 1.01; 95% CI of 0.97–1.05, p = 0.54). Similarly, presence of sensory dysfunction did not predict anisotropic bias.

Primary analysis: generalised Tactile anisotropy

Descriptive analysis of response frequency supports the presence of tactile anisotropy occurring bilaterally in all participants. That is, across-oriented stimulus pairs were perceived as further apart independent of whether there is a factual difference in width-distance between stimulus-pairs and there were no differences between hands (χ²(1): 0.004, p = 0.948). The proportion of across-oriented stimulus pairs being perceived as further apart on each hand are listed in Table 4 and the general predictor set for tactile anisotropy to occur are listed in the supplementary section Table 2.

General set of predictors for all participants

Tactile anisotropy did not differ as a function of hand (OR 1.040, 95%CI [0.883–1.226], p = 0.637), but some groups differences were present. In the multivariate model of best fit (which included gender, length difference of stimulus pairs, current and average pain intensity), the PC group had reduced odds of perceiving across-oriented stimulus pairs as further apart than pain-free controls (OR: 0.632, CI 95% [0.407–0.981], p = 0.041; see Table 2 in the supplementary section). People with CRPS had no difference in the odds of perceiving across-stimuli as farther apart than pain-free controls (OR: 0.623, 95% CI [0.362–1.074], p = 0.088). While approaching statistical significance, the wide confidence intervals suggest large variability of any effect.

Patient specific clinical variable set

Considering both pain groups, results show that a combination of illness duration, average pain intensity, and stimulus pair distance difference (0/1/2 cm) best predicts the tactile anisotropy on a multivariable level. Similar to above, hand perception did not predict tactile anisotropy—it was not included in the multivariate model (did not add to model fit), and univariate analyses were not statistically significant (Body perception scale OR: 1.10; 95% CI of 0.809–1.495, p = 0.544).

To further explore the anisotropy data, the GEE models using group x hand and group x hand x stimulus pair length difference interactions were used to create psychophysical curves of tactile distance judgements on both hands for all three groups (see Fig. 3). Visual inspection suggests that the slope increase of the curve for the affected hand of the CRPS group may differ from that of the pain-free control group but this trend does not meet statistical significance. Table 5 provides the PSE values for tactile distance judgements as function of hand and group. Here, PSE represents the point at which the distances between across- and along-oriented stimulus pairs are perceived as subjectively equal. These PSE values are largely similar between groups (i.e., we did not detect a statistically significant difference between the groups).

Secondary analyses: Accuracy of tactile distance perception (orientation independent and orientation dependent)

The GEEs evaluating the tactile distance accuracy outcomes confirmed that all groups were equally accurate in discerning the longer length in both directions on both the affected (dominant) and non-affected (non-dominant) hand. Importantly, neither clinical nor sensory signs, nor distorted hand perception, were found to influence the accuracy of these tactile distance judgements (neither orientation-dependent nor independent). See the supplementary tables for the proportion of accurately identified stimulus-pairs in both directions and the general and patient-specific predictors for accurate tactile distance perception set.

Anisotropic perception bias, and tactile anisotropy

One argument for the lack of magnified anisotropic perception bias (and tactile anisotropy) in people with CRPS is that they may merely be inaccurate at the task, and thus, consequently increase noise in the data renders a between-group signal undetectable. However, CRPS participants demonstrated correct recognition of width differences (independent of orientation; see supplementary analysis, see secondary outcome one), which suggests that our results were not due to discrimination errors. Bilaterally preserved accuracy—despite allodynia, pain or hypaesthesia—also suggests that receptive field and receptor properties on the hand’s dorsum (and centrally) are not affected by CRPS pathophysiology. Similarly, that we did not detect a difference in anisotropy between hands in those with unilateral CRPS or unilateral hand pain (PC) suggests against a role of dysfunctional spatial processing of tactile input in contributing to hypoaesthetic signs or distorted hand perception, both of which were observed in the affected, but not unaffected, hands.

These results would not be predicted based on what is currently understood regarding the S1 representation of the affected and non-affected hand in CRPS or the contemporary knowledge of composition of S1 hand representation (Azanõn et al., 2016; Di Pietro et al., 2015; Di Pietro et al., 2016; Gallace & Spence, 2010; Gallagher, 2005; Medina & Coslett, 2010; Serino & Haggard, 2010). That is, given previous observed between-hand differences in S1 hand representation in unilateral CRPS, we would expect anisotropic differences as well. Here it is relevant to consider the underlying proposed mechanisms of anisotropy. While it is unclear how central and peripheral anisotropies are linked, tactile anisotropy is thought to constitute a general operating principle of body representation (Green, 1982; Knight Fle, Longo & Bremner, 2014; Longo, Ghosh & Yahya, 2015; Miller, Longo & Saygin, 2016).

For S1 hand representation, it is thought that the numbers of peripherally stimulated oval-shaped RFs are ‘counted’ cortically, and since the number of stimulated oval-shaped RFs on the hand’s dorsum is higher in medio-lateral compared to proximo-distal orientation, limbs are represented ‘wider and squatter’ than they really are (Longo & Haggard, 2011). It is thought that this ‘fat’ bias of orientation-specific distance perception mirrors homuncular distortion of S1 hand representation (and thus may influence perceptions and cognitions of the hand) (Longo & Haggard, 2011; Longo & Haggard, 2012; Miller, Longo & Saygin, 2016).

Drawing on the proposed mechanisms underlying tactile anisotropy, it may therefore be possible that our results differ from past work because S1 morphology is actually disrupted in a different way than currently presumed for CRPS. Two limitations of the neuroimaging evidence for altered S1 representation in CRPS are relevant here. First, most studies are compromised by poor blinding, incomplete reporting or lack of control group or control hand (Di Pietro et al., 2013) which suggests caution when interpreting many CRPS neuroimaging results. Second, all studies have relied on Euclidean distance between thumb/index finger, and the fifth finger, which does not take into account that the cortex is curved, not flat (Baliki et al., 2011). A recent study that overcame this limitation demonstrated no abnormalities in digit-specific organisation of S1 in people with CRPS (Mancini et al., 2019) which seems consistent with the current findings and point to a lesser, or different, role of S1 changes in CRPS than has been widely accepted.

Notwithstanding challenges in imaging, it is certainly possible that tactile anisotropy might be preserved given that processing of other types of somatosensory input have been shown to be intact in CRPS. Indeed, while our results were surprising, past work has demonstrated that aspects of somatosensory processing and multisensory integration are preserved in people with CRPS—i.e., functions that involve neural mechanisms distinct from those of static TPD measures (Moseley & Wiech, 2009; Reinersmann et al., 2013; Reiswich et al., 2012). Such findings suggest that the somatosensory processes underlying measures of tactile distance judgements and TPD differ and are consistent with suggestions that TPD measures likely reflect a deficit of processing intensity cues rather than spatial cues in CRPS or a disruption in fine-grained maps of digits on the CRPS-affected hand (Bruns et al., 2014; Craig & Johnson, 2000; Craig & Kisner, 1998; Mancini et al., 2013; Tong, Mao & Goldreich, 2013).

Distorted hand perception despite intact somatosensory spatial processing of tactile input

While spatial processing of tactile input was preserved, hand perception was distorted in CRPS and unrelated to anisotropic perception bias (i.e., the presumed indicator of the ‘fat’ S1 hand representation). Specifically, the outcome on the task measuring hand representation (tactile distance judgements) was unrelated to all measures of hand perception (visual and cognitive-affective hand perception, measured with template—matching and questionnaire).

Our results are generally consistent with both the lack of correlation between different body representation outcome measures (Longo, 2015) and the observed distorted hand perception in CRPS (Frettloh, Huppe & Maier, 2006; Galer & Jensen, 1999; Lewis & McCabe, 2010). Body representation research often finds that outcomes of different measures are unrelated and suggests that this reflects the continuum along which the body is represented in different dimensions (somatosensory, visual, cognitive/affective etc.). On this continuum, top-down and bottom-up composed representations reference onto each other for any given task outcome (Gallagher, 2005; Longo, 2015; Longo & Haggard, 2012). The complexity of this interaction or reciprocal referencing may then also explain deficits in some types of body representation while other representations (such as S1 hand representation) are intact even within the same sensory modality (i.e., tactile).

This notion of complex representation is also relevant to the composition process of the percept itself: tactile stimulation starts a complex process of sensation that typically ends with a percept—one feels the touch at a specific location on the body. The percept results from bottom-up processing (tactile sensation) and top-down interpretation of these sensations against a stored inner S1 hand representation. A percept thus integrates a multiplicity of factors: receptor density in the skin, S1 receptor field geometry and orientation-selective neurons of receptor fields, cortical magnification and perceived hand size. Moreover, there are multifactorial influences on every type of body representation, whereby top-down interpretations also incorporate psychosocial & environmental imprints and reflect ‘mentalisation’ of sensorimotor signals (Fotopoulou & Tsakiris, 2017). Given all these proposed contributions to a percept, it is possible that some aspects are more heavily weighted in certain tasks than others and thus may well result in differences between body-representation outcomes. The ultimate percept may then simply reflect the reciprocal referencing of body representations that varies as a function of task demand or other factors.

Relationship between S1 hand representation and hand perception

Despite the fact that tactile anisotropy exists (‘fat, squat’ S1 hand representation, (Longo & Haggard, 2011), we generally do not explicitly perceive our hands as fat and squat. It is intriguing that the phenomenology of CRPS distorted hand size perception appears to parallel what would be predicted with tactile input—and its anisotropic property—alone. That is, in healthy controls and people with non-CRPS hand pain, tactile distance perception is integrated into our conscious hand perception in a way that results in an accurate sense of the size and shape of our hand. However, in CRPS, tactile distance perception is normal and conscious hand perception is not. Is it possible that, in CRPS, the usual weighting of tactile distance perception and S1 hand representation is disrupted (see Fig. 4) and that the perceptual distortion reflects altered weighting of somatosensory input in the production of conscious hand perception?

This speculative idea has some backing in the literature. For example, in anorexia nervosa, there is greater than usual connectivity within somatosensory cortex and less than usual activity within areas associated with the visual processing of faces and bodies (extrastriate body area and fusiform area) (Kim et al., 2017) . This has led to the proposal that body perception disturbances in that group reflect the relatively greater weighting of somatosensory (tactile) body representations (i.e., wider, squatter) (Longo, 2015). Furthermore, although an increased weighting of somatosensory representation might not conclusively explain anorectic perceptions (Green, 1982), ‘swollen or fat’ perceptions have been observed under conditions of neuronal or sensorimotor disruptions. For example, local anaesthetic to the finger results in a perception of a ‘fat’ finger, with proprioceptive processing in S1 appearing to have a role in this distortion (Gandevia & Phegan, 1999; Lackner, 1988; Moseley et al., 2006). Indeed, mismatches between proprioception and other sensory/motor inputs have repeatedly been discussed in relation to altered CRPS body size/shape perceptions—and results of the present hand questionnaires match these discussions (Lackner, 1988; Lewis et al., 2010; Moseley, 2004; Moseley et al., 2006; Reinersmann et al., 2010; Reinersmann et al., 2012).

Together such findings suggest that while spatial processing of tactile input and S1 hand representation appear preserved (as our results suggest), disruptions in other sensory processing networks might be present and account for a ‘altered weighting process’. One possibility is that altered or increased functional connectivity of somatosensory and motor cortex (S1/M1 sensorimotor cortex) to the right insular and other motor areas might impact on a weighting process of somatosensory regions—in terms of amplified weighing—and that this might contribute or result in an ‘bizarre and foreign’ hand perception. Imaging studies for example show alterations in functional connectivity between regions of the motor and intraparietal sulcus (IPS) that serve as an interface when integrating perceptual and motor information (Bolwerk, Seifert & Maihofner, 2013; Kim et al., 2017; Maihofner, Handwerker & Birklein, 2006; Pleger et al., 2014). Additionally, there is evidence in CRPS of reduced resting state functional connectivity between S1 and insula regions which show decreased endogenous inhibitory control of sensory processing regions in the insula (Kim et al., 2017).

Finally, we acknowledge that just as in anorectic body perception the distorted CRPS hand perception may reflect an increased weighting of top-down (cognitive-affective) interpretations rather than a (cortically driven) over-weighting of somatosensory (tactile) hand representation. In the context of aversive and ambiguous sensation (inexplicable pain, reduced motor function, decreased awareness of hand position, swollen feelings) and an unaesthetic appearance, the CRPS sufferer may perceive the hand as foreign and disturbing. There are some self-report data that suggest this to be the case (Lewis & Schweinhardt, 2012; Michal et al., 2017; Moseley & Butler, 2017).

Experiencing symptoms that even a pain specialist cannot explain may reinforce rejection of the hand or difficulties to integrate it into the own body concept. This in turn opens the option that psychoeducational interventions, and/or a combination of psychoeducational measures and cognitive behavioral therapy might impact on the functionality of mutually operating S1 hand representations. There is some preliminary support for this idea with observational data clearly suggesting better impacts on bodily perception from a psychoeducational approach tailored to CRPS than from a similar untailored approach (Moseley & Butler, 2017).

While this is the first study to use robust measures of spatial processing of tactile input in people with CRPS, and concurrently measure hand perception, a limitation of our study is that we did not include a neuroimaging component which would have allowed formal exploration of concurrent cortical changes. Future research linking neural and perceptual measures of the hand with evaluation of numerous features of spatial acuity is clearly warranted to delineate tactile dysfunction in people with CRPS. A second limitation is that we did not formally lodge a detailed protocol prior to data collection or analysis, a procedure now recommended in pain studies to promote accountability, transparency and openness (Lee et al., 2018). We did develop an a priori experimental protocol that followed a standardized procedure in the use of the new paradigm and use of well-established questionnaires, as described in the methods section. Additionally, our a priori sample size calculation was powered to detect a small to moderate effect for the interaction between Group and Hand using a planned RM ANOVA. We instead used GEE to analyse the data, meaning our calculation was not specific to our analysis; but, as mentioned GEE analyses tend to have higher power with lower sample numbers than RM ANOVAs (Ma, Mazumdar & Memtsoudis, 2012), which suggests this may not have been a problem. It is relevant to note that the confidence intervals for tactile distance judgements in the CRPS group were very wide, suggesting that any between-group difference is likely small. This may suggest that our initial choice of powering for a small-to-moderate effect size may be too liberal. Regardless, even if significant differences were present between groups, our findings for anisotropic perception bias and tactile anisotropy were in the opposite direction than hypothesized. Last, our hand perception task used standardized hand images, meaning that between group differences in perceived hand size could reflect between group differences in actual hand size. However, individual differences in actual limb size are typically greater than group differences, suggesting against a systematic difference underlying the present results. That this task has been used in this way clinically and in pilot trials in our group, and between group differences between those with CRPS and pain-matched controls, supports its use here (Moseley, 2005). It does remain possible that group differences represent not a difference in perceived hand size but rather differences in the ability of an individual to transpose their own limb to the pictured limbs, but even then, we would expect increased unsystematic error. That body perception questionnaire responses about limb size typically matched the visual scaling task choices suggests that it does capture perceived hand size.

Our study did also have important strengths. These include the use of a paradigm that incorporates the orientation-dependence in spatial processing of tactile input, the use of a non-CRPS pain patient control group, as well as the use of GEEs to evaluate the data set which reduces data noise for interrelatedness of cells in repeated measures and the variability associated with heterogeneity of chronic pain. Further, use of comprehensive analyses to also consider tactile distance judgement task accuracy ensures our results of maintained anisotropic perception bias and general tactile anisotropy are not due to task error.