Cohort selection

Medical records from a 2 year period (March 2013–2015) at Fitzpatrick Referrals Orthopaedics and Neurology were searched for CKCS that had undergone MRI (119 dogs). The medical records were analysed by CR. Ten dogs were excluded because the imaging did not include the brain and cervical spinal cord, 1 dog was excluded because of a diagnosis of central nervous system inflammatory disease and 19 dogs were excluded because of incomplete medical records, diagnosis of skin disease and/or the cause of pain or scratching was equivocal, and/or because the dogs had not been examined by an ECVN diplomat/resident. This included 16 CKCS that were presented for health screening prior to breeding. The remaining 89 dogs were divided into phenotypical groups according to clinical and SM status and 49 dogs without clinical signs associated with SM were further excluded leaving two groups: 1) SM with phantom scratching (19 dogs) 2) dogs with clinical signs of SM including behavioural signs of pain but no phantom scratching (21 dogs). For the purposes of this study, SM was defined as a fluid filled cavity equal to or greater than 2 mm in diameter associated with CM detected on MRI. The presence of phantom scratching was defined as a repetitive scratching action towards the neck region which could be induced by stimulation of a receptive field in the cervical region and/or induced by excitement or exercise in the absence of generalised pruritus and when clinical examination did not suggest skin and external ear disease (for example skin reddening or crusting) and when the scratching behaviour had been observed by an ECVN diplomate or resident. After review of the MRI, three cases were subsequently excluded as their MRI did not include transverse imaging of the cervical spinal cord. The remaining total study cohort therefore comprised 19 dogs with phantom scratching (“scratchers”) and 18 dogs with SM but no phantom scratching (“non-scratchers”).

MRI examination

All MRI examinations were performed on 1.5 T scanner (Symphony Maestro Class, Siemens, Enlargen, Germany) with dogs under general anaesthetic. All dogs were positioned head first in dorsal recumbency and a combination of head and spine coil was used for all studies. The studies included routine T2-weighted MRI sequences in sagittal and transverse plane of the cranium and spinal cord. Only images from the cervical spinal cord were included in the study. The transverse sequences were centred on the widest part of a syrinx and covered the entire cervical syrinx. The study parameters for the sagittal plane were TR 2700 ms, TE 104 ms, FOV 220 mm, Slice thickness 2 mm, Matrix 512 × 512, whereas parameters for the transverse plane were TR 3200 ms, TE 109 ms, FOV 130 mm, Slice thickness 3–3.5 mm, Matrix 384 × 384.

Measurements

The MRI studies were anonymised and randomised by JJ who was blinded to the clinical status. The MRI studies were viewed by DICOM viewing software eFILM workstation (Merge Healthcare 900 Walnut Ridge Drive, Hartland, WI 53029 USA).

Investigation 1: Is phantom scratching associated with a large dorsolateral syrinx in a particular region of the cervical spinal cord i.e. certain cord segments?

For each transverse MRI slice in cervical area, presence or absence of a syrinx was recorded. If present, the extent of the syrinx within the spinal cord was determined by measuring the maximum perpendicular dimensions of the cavity in each spinal cord dorsal quadrant in each transverse MRI section (Fig. 2). After measurements were completed the maximum dimensions of the syrinx (in cm) in each dorsal quadrant in each transverse MRI section were multiplied to give a value that reflected the perpendicular dimensional area of the cavity in each quadrant for each transverse MRI section. Subsequently a figure representing the dorsal syrinx cavity size for each spinal cord segment was obtained by calculating the mean of all the perpendicular dimension areas for each transverse MRI section for each segment for each dog.

Statistics

Measurement reliability was assessed for inter and intra observer reliability using intra-class correlation [ICC] model [1, 2] [18] and the kappa statistic where appropriate. For Investigation 1 quadrant areas were summarised using mean, standard deviation, median along with minimum/maximum. These values were calculated from all the measurements from each transverse MRI section for each dog. Due to the non-Normality of the data, comparisons between scratchers and non-scratchers were made using Mann–Whitney tests at a 5% level of significance. In Investigation 2, cross-tabulations of Group by whether syrinx was extended to the SDH or not were constructed for each spinal segment and, where appropriate, chi-square tests were used to assess any associations. Fisher’s exact tests were then used to test differences in the proportion of dogs within each group with an extended syrinx.

Investigation 1. Is phantom scratching associated with a large dorsolateral syrinx in a particular region of the cervical spinal cord i.e. certain cord segments?

The study found that there was greater dorsal quadrant involvement in the scratcher group for spinal cord segments C3 – C6 (vertebrae C2 – C5). This suggested that phantom scratching may be associated with a large dorsolateral syrinx in the mid cervical spinal cord and illustrated by boxplots (Fig. 3). 100% of the 19 scratchers had large syringes in the dorsal quadrants of spinal segment C4 (vertebrae C3) indicating that this might be a particular region of interest in future studies. (Fig. 4). Spinal segments C1, C2, C7 and C8 had low numbers of dogs and therefore either could not be tested statistically or the results should be viewed with caution.

Investigation 2. Is phantom scratching associated with a dorsolateral syrinx that extends to the SDH?

Scratchers were significantly more likely to have SDH involvement in the C3 to C5 spinal segments (C2 to C4 vertebrae) and less so for the C5 spinal segments (C6 vertebra). There was a marginal significant difference for the C6 spinal segment (C7 vertebrae) and no significant difference at C7. The number of dogs that had a syrinx extending to the SDH was low, once again, for spinal cord segments C6 and C7. Spinal cord segments C1 and C8 did not have a syrinx in any dog. In addition scratchers were more likely to have SDH involvement over several transverse slices. The mean ± standard deviation and median number of affected slices was respectively 10.8 ± 6.8 and 10 (range 0–28) for scratchers and 0.6 ± 1.7 and 0 (range 0–7) for non-scratchers.

What is the neuroanatomical site pivotal for the phenomenon of phantom scratching?

This study suggested that phantom scratching in the dog may be associated with a large dorsolateral syrinx in the C3-C6 spinal segments (C2-C5 vertebrae). This correlated with the region of syrinx extension to the SDH (Fig. 4). An earlier hypothesis by this group was that the first two cervical segments may be pivotal for phantom scratch. However this study found low numbers of dogs with syringes at this site and the hypothesis was not proved. The first two cervical segments were of interest for several reasons. First, they contain the lateral cervical nucleus (relay for ascending spinocervicothalamic tract fibres). Second, a model of fictive scratch can be created in the decerebrate/spinal cat by application of tubocurarine to the dorsal surface of the cervical cord at C1 (and to a lesser extent C2) and then rubbing the pinna and the skin behind the ear [19, 20].

Scratchers were also likely to have SDH involvement over many adjacent transverse MRI slices and therefore cervical spinal cord segments. This may be important because it is suggested that lamina I of the SDH is a functionally interconnected layer with local circuit neurons connecting the dorsal grey matter of several spinal cord segments [21]. From this study it might be concluded that a dog with MRI evidence of a syrinx extending to the SDH is at risk for phantom scratching. Conversely, if the syrinx in a scratching dog with SM does not extend to the SDH and/or there is not extensive dorsolateral spinal cord involvement, then another explanation for scratching should be sought before a diagnosis of SM associated scratching is made. In the authors’ experience, presentation with a history of caudal head and/or ear scratching/rubbing is not unusual with brachycephalic toy breed dogs with suspected CM-pain [2, 5]. However the nature of the scratching is different in such cases. In the authors’ experience, dogs with suspected CM-pain may scratch to the sides of the head/ears but make skin contact and may vocalise during or shortly before scratching (with ear disease excluded as a cause of these signs). Since the phenomenon of SM phantom scratching has been publicised by social media, there can be a risk of misdiagnosis. Some dogs with SM are asymptomatic and there are many alternative explanations for scratching, especially if the pruritus is generalised or of multiple digits.

Neuronal population and neurotransmitters within the C3-C6 SDH neuroanatomical site which could be damaged or modified by a syrinx

Activation of a population of gastrin-releasing peptide receptor–expressing interneurons in the SDH is key in itch transmission [22, 23]. These pruritoceptive circuits are regulated by GABAergic inhibitory dorsal horn interneurons [22, 24]. Glutamate release is pivotal for itch modulation [25] which may explain why (anecdotally) gabapentin and pregabalin suppress SM phantom scratching, drugs commonly prescribed for this purpose. Other paramount molecular signalling include substance P, calcitonin-gene related peptide and serotonin [9]. Substance P binds to the tachykinin neurokinin 1 receptor (NK- 1R), Recent studies have indicated that although gastrin-releasing peptide receptor–expressing neurons contribute to hyperknesis, it is the NK-1R expressing neurons located in the dorsomedial aspect of the SDH that are pivotal to alloknesis or ongoing itch [26]. These NK-1R expressing neurones give rise to ascending somatosensory projections [26] and represent a potential target to treat chronic itch [27]. A NK-1R antagonist, maropitant citrate, can reduce the size of lesions in C57BL/6 mice with idiopathic ulcerative dermatitis, a rodent model of chronic itch, which is characterised by lesions of the neck and shoulders and scratching behaviour induced by substance P [28]. Nonetheless, although NK-1P mediated itch seems a plausible theory for phantom scratching, there is the paradox that phantom scratching seems to be associated with damage to the SDH and if this were true then this neuronal population may be compromised. Also there is an argument that phantom scratching is not the same as neuropathic itch because there is no skin trauma. It is therefore more logical to hypothesise that phantom scratching may be due to damage to an inhibitory GABA and/or glycine neuron population. In the spinal cord dorsal horn, 30–40% of all neurons that regulate the transmission of information that is eventually perceived as touch, pain and itch are inhibitory [29]. These inhibitory neurons are found in all laminae except laminae IX and in rodents, GABA and glycine expressing interneurons account for appropriate a quarter of the neuronal population in the SDH [30, 31]. Recent studies have identified a pivotal dorsal horn neuronal population which express neuropeptide Y and inhibitory interneuron markers [32, 33]. Loss of these neurons produces disinhibition of a presumed and unknown excitatory neuron. Importantly, this neuronal population is activated by mechanical stimulation and is independent from puritoceptor and histamine transmission [32, 33] i.e. represent a possible candidate cell for phantom scratching which is also triggered by mechanical stimulation. The next stage in this study is compare neuronal populations in histological sections from CKCS with syringomyelia in order to elucidate this hypothesis further.

Can we better understand the mechanism of scratch and the scratching central pattern generators (CPG)?

Scratching is a stereotyped sequence of muscle contractions that may be produced in a reliable but flexible manner. These, and other regular movements such as locomotion, are controlled by local neural circuits or CPG that are capable of producing repetitive movements, even when isolated from the brain and sensory inputs [13, 34, 35]. Experiments have suggested that the CPG for pelvic limb scratching is in the lumbosacral segments [36]. Therefore, if phantom scratching was similar to fictive scratch, it would suggest that the syrinx damages a neuronal population which influences the descending projections to the lumbosacral CPG for scratching. It is hypothesised that a logical explanation could be loss of an inhibitory interneuron population and this results in a disinhibition of scratching neural circuits and (presumed) synaptic reorganisation. The link between the cervical dorsal horn and lumbosacral intumescence in fictive scratching is the propriospinal system [37]. Propriospinal neurons are interneurons which connect multiple spinal cord segments and participate in complex or “long” motor reflexes including modulating CPGs [38, 39]. They have considerable plasticity and are pivotal in functional recovery after incomplete spinal cord injury [39]. Long propriospinal neurons have cell bodies in the cervical dorsal horn and the axons course caudally both in the superficial layers of white matter in the ventral and the lateral funiculi to terminate in the lumbosacral grey matter [39]. Intriguingly spinal segments C3 and C4 are the site of a population of propriospinal neurons termed the C3-C4 propriospinal system. This is pivotal for control of forelimb voluntary movement which connect to propriospinal neurons that project to more caudal spinal segments [39, 40]. Given the site, it is tempting to consider this might be involved in SM phantom scratching however this is conjecture. Finally the role of C-bouton synapse (spinal cholinergic interneurons) in fictive locomotion (spinal walking) has recently been demonstrated [41]. These neurons original from interneurons located lateral to the central canal and are responsible for increased motor neuron excitability in fictive locomotion [41, 42]. As a potential natural model of fictive scratch, dogs with SM associated phantom scratching may represent a huge resource for understanding neural plasticity and further study is warranted.

Study limitations

This study had limitations of which the reader should take account. The number of study animals was small. Furthermore our MRI protocol dictates that a block of transverse images are planned to cover the syrinx so that there is a “slice” perpendicular to the spinal cord through the widest part of the syrinx. This method ensures that the extent of the syrinx can be appreciated but meant that the exact position of the transverse slices varied between dogs. It also meant that syrinx free portions of the spinal cord were less likely to have transverse images. Every effort was made to compare like with like in terms of slices but perfect matching was inevitably not possible however the differences were considered to be occurring at random and the same frequency in both groups. Consequently measurements of the length of syrinx grey column involvement would have been inaccurate and therefore were not obtained. A further limitation was that measurements of thoracic and lumbar syringes were not included because of the inconsistency of area and completeness of the transverse imaging caudal to the cervical region. Therefore the possibility of more caudal influences on phantom scratching were not investigated and it should be considered that the association of a wide mid-cervical syrinx and phantom scratching may be a coincidence especially as the mid cervical region is the most common site for syrinx development [43]. A further limitation of the study was the definition of SDH involvement as extension of the syrinx to the dorsolateral margin of the spinal cord so that that rim of spinal cord was not or only just appreciable could be regarded as subjective and inaccurate. The “rule” that was followed during the study was that there was SDH involvement if the fluid in the syrinx and the subarachnoid space looked confluent. The smallest unit of measurement of eFILM is 1 mm so a more objective measure was not possible.