POEMS syndrome—an acronym for polyneuropathy, organomegaly, endocrinopathy, monoclonal gammopathy, and skin changes—is a unique multisystem disorder that is also known as Crow–Fukase syndrome.1 Because it is strongly associated with plasma-cell dyscrasia, especially osteosclerotic myeloma, monoclonal proliferation of plasma cells has been thought to play an important role in the development of various POEMS symptoms.1 2 In addition, serum concentrations of cytokines, such as interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumour necrosis factor-α (TNF-α), are elevated.3 Recent studies implicate vascular endothelial growth factor (VEGF) as another pathogenetic factor.4 5 As this syndrome manifests a variety of symptoms, it has been typically examined from the viewpoint of wide-ranging organ involvement;1 2 6 7 reports exclusively examining its neuropathic features are relatively rare. Furthermore, because electrophysiological and histopathological features are those of demyelinating neuropathy,7 8 neuropathy in POEMS syndrome has often been diagnosed as chronic inflammatory demyelinating polyradiculoneuropathy (CIDP) when the associated symptoms other than neuropathy are not conspicuous.9 10 New therapeutic approaches to POEMS syndrome have been described that differ from CIDP approaches.6 10^–13 Therefore, clarification of the clinical and pathological features and their correlations unique to POEMS syndrome is needed to improve the diagnostic accuracy and subsequent therapeutic consequences.

The present study describes the clinical, pathological and cytokine profile features that are unique to POEMS syndrome.

PATIENTS AND METHODS

Patients

Twenty-two consecutive patients with POEMS syndrome who were referred to the Department of Neurology of Nagoya University Hospital from 1987 to 2007 were investigated. Patients included 13 men and 9 women aged 54.8±13.6 (mean ± SD) years. Clinical features and laboratory data assessed before the initiation of treatment are summarised in table 1. As for the assessment of neuropathic pain, patients were asked whether they have spontaneous pain. Mechanical stimuli, both normally painful and normally non-painful, were applied on the distal portion of the lower limbs (dorsum of foot and lateral surface of the lower leg) by examiners to evaluate hyperalgesia. The patient’s subjective responses to these stimuli were qualified and the amount of pain was described in categorical scale (none, mild, moderate and severe). Because patients with POEMS syndrome lack a contralateral homologous normal side due to their symmetrical polyneuropathy pattern, responses to normally painful stimuli such as pin-prick and pinwheel were compared with those on a more proximal portion of the limbs (anterior thigh). Abnormal sensations without pain were not considered to be hyperalgesia. Patients’ functional status was assessed at the peak phase according to the modified Rankin Scale.14 Motor and sensory conduction was measured in the median, tibial and sural nerves in all patients during their initial clinical assessment by neurologists, using a standard method with surface electrodes for stimulation and recording.

Pathological assessment of sural nerve biopsy specimens

Sural nerve biopsy was performed in 21 patients before the initiation of treatment as described previously.15^–17 Specimens were divided into two portions. The first was fixed in 2.5% glutaraldehyde in 0.125 M cacodylate buffer (pH 7.4) and embedded in epoxy resin for morphometric and ultrastructural study. Densities of small and large myelinated fibres were assessed in toluidine-blue-stained semithin sections using a computer-assisted image analyser (Luzex FS; Nikon, Tokyo, Japan), as described previously.15 16 18 To determine the extent of nerve fibre loss, the total number of nerve fibres in complete transverse sections of sural nerves was estimated because the density might be reduced without actual nerve fibre loss when the endoneurium was enlarged by a pathological condition such as endoneurial oedema. Therefore, the total endoneurial area and subperineurial space devoid of nerve fibres were assessed using the image analyser (Luzex FS) in cases in which complete transverse sections of the sural nerve could be obtained; we estimated the total number of nerve fibres in these cases. Two cases (patients 9 and 14) were excluded because their sural nerve was only partially obtained. To determine the total numbers per complete cross-section of the sural nerve, the density of each morphometric index was multiplied by the endoneurial area from which the subperineurial space devoid of nerve fibres was subtracted.17 The remainder of the glutaraldehyde-fixed sample was processed for the teased-fibre study, in which at least 200 single fibres were isolated and their pathological condition was assessed microscopically according to criteria described previously.15 19 Each teased fibre was divided into two categories based on the diameter of the middle portion of the largest segment. Fibres of 6.73 μm or more in diameter were designated as large fibres and those of less than 6.73 μm as small fibres.15 Fibres showing axonal degeneration were not included when the diameter of the fibres was evaluated. The second portion of each specimen was fixed in 10% formalin solution and embedded in paraffin. Sections were cut by routine methods and stained with hematoxylin and eosin, Congo red, the Klüver–Barrera method and the Masson trichrome method.

For the electron microscopic study, epoxy resin-embedded specimens were cut into ultrathin transverse sections and stained with uranyl acetate and lead citrate. To assess the density of unmyelinated fibres, electron microscopic photographs were taken at a magnification of 4000× in a random fashion to cover the area of ultrathin sections, as described previously.15 16 20 These electron micrographs were enlarged to about 6500×. The total area analysed in electron micrographs was at least 0.03 mm², obtained from at least three fascicles. Unmyelinated fibres were distinguished from Schwann cell cytoplasmic profiles by their round or oval shape, a lighter appearance than Schwann cell cytoplasm, and often a higher incidence of microtubules.21 The presence of mesaxon-like structures and a greater density of axolemma than Schwann cell membranes were also criteria for identifying unmyelinated fibres.21 Disproportionately large unmyelinated fibres over 3 μm in diameter were not counted because these fibres were considered to be originally myelinated fibres and formed as a consequence of demyelination.17 22 A conglomerate of Schwann cell processes with or without unmyelinated axons, and enclosed by a continuous loop of basal lamina, was designated as a “Schwann cell subunit”, as described previously.17 22 Only Schwann cell subunits related to unmyelinated fibres were counted. Unmyelinated fibres found in Schwann cell subunits that previously contained myelinated axons (ie, bands of Büngner) were not counted because they could be sprouts from regenerating myelinated fibres.21 23 In addition, unmyelinated fibres and Schwann cell subunits that took part in the formation of regenerating clusters of myelinated axons were not counted for the same reason.22 The bands of Büngner were distinguished from subunits of non-myelinating Schwann cells, as described previously.17 21 22

The widening of major dense lines of myelin lamellae was designated as uncompacted of myelin. Myelinated fibres were considered as presenting uncompacted myelin lamellae (UML) when at least three lamellae were not compacted along a semi-circumference of the myelin sheath or in an additional loop.24 The percentage of fibres with UML in the large myelinated fibres and small myelinated fibres were separately assessed. The area of ultrathin section was overviewed by photographs at a magnification of 2000×. Fibres with UML confirmed by greater magnification were marked on the photographs and the diameter of each myelinated fibre was assessed using a computer-assisted image analyser (WinROOF; Mitani, Fukui, Japan). At least 300 fibres were assessed to determine the frequency of UML.

Control values were obtained from five autopsy cases in which the patients died of non-neurological diseases (male:female, 1: 4; age range: 35–71 years, mean ± SD: 55.8±14.5 years). Specimens were processed in the same manner as for POEMS syndrome patients.

Assessment of serum cytokines and cytokine receptors levels

All patients were examined before initiation of treatment. Measurements of IL-1β, IL-6 and TNF-α were performed in 15 patients and values were compared to 20 control patients with other neurological diseases, including spinocerebellar ataxia, multiple system atrophy and spinobulbar muscular atrophy. Measurements of IL-1 receptor antagonist (IL-1ra), soluble IL-6 receptor (sIL-6r), soluble TNF receptor I (sTFNrI) and soluble TNF receptor II (sTNFrII) were performed in 12 patients and values were compared to those of 7 patients with other neurological diseases. Peripheral blood was taken in dry tubes and centrifuged within 3 hours. The sera were kept frozen at −80°C until analysed. Blood samples were obtained in the absence of overt fever, infection and shock. An enzyme-linked immunosorbent assay kit was used for the quantification of IL-1β (BioSourse, Nivelles, Belgium), TNF-α (JIMRO, Takasaki, Japan), IL-1ra (R&D Systems, Minneapolis, USA), sIL-6r (R&D Systems), sTFNrI (R&D Systems) and sTFNrII (R&D Systems), and a chemiluminescent enzyme immunoassay kit was used for the quantification of IL-6 (Fujirebio, Tokyo, Japan).

Statistical analyses

Quantitative data, presented as the mean ± SD, were compared with control values. Statistical analyses were performed using χ² test or the Mann–Whitney U test as appropriate. Values of p less than 0.05 were considered to indicate significance.

RESULTS

Neuropathical features

Neuropathic features are summarised in table 2. All patients showed a symmetrical polyneuropathy pattern with greater involvement of the lower rather than the upper limbs. Sensory symptoms were moderate to severe in most patients in the distal portion of the lower extremities as a whole. Nineteen patients (86%) complained of numbness in the lower extremities, most of which were uncomfortable sensations, such as aching, tingling or prickling. Twelve patients (55%) reported spontaneous pain. Painful symptoms generally were brought on, or made worse, by gentle manual pressure or pinching. Many of the patients complained of pain as they walked because pressure was applied on the soles of their feet. Patients also reported pain when a blanket was put on their feet. Distally accentuated hyperalgesia to pin-prick or pinwheel was also observed. As a whole, 14 patients (64%) reported hyperalgesia, as described above. In some patients, pain was the most characteristic feature, significantly compromising activities of daily living. Seven patients (32%) reported having difficulty walking due to pain (patients 1, 4, 9, 10, 15, 18 and 21). On the other hand, three patients (14%) did not complain of any positive sensory symptoms despite the presence of subjective sensory deficit (patients 8, 12 and 14). Autonomic symptoms were reported in 4 patients (18%): these consisted of impairment of sweating in the extremities in patients 8, 13 and 15; urinary retention in patient 13; and constipation and orthostatic hypotension defined as a fall of 20 mm Hg in systolic blood pressure following arising from the supine position in patient 18. In addition, 6 of 13 male patients (patients 5, 10, 15, 16, 19 and 22) complained of impotence, although this may have been related to endocrine disorder.

Nerve conduction studies revealed slowing of motor and sensory conduction velocities and prolongation of distal latencies in all patients, as shown in Supplementary Material 1.

No significant difference in the type of M-protein, duration of neuropathy, or relative predominance of weakness and sensory deficit was present between the group with mechanical hyperalgesia and that without it. Functional status assessed by the modified Rankin Scale was not significantly different between the two groups. Electrophysiological features were different only in the amplitude of sensory nerve action potentials of the sural nerve, which was more profoundly reduced in the group with hyperalgesia (p<0.05).

Pathological findings of sural nerve biopsy specimens and their correlation to neuropathic pain

Quantitative data, including those of nerve fibre density of individual cases, are listed online in Supplementary Material 2. As for myelinated fibres, the density varied from 1396 to 6558 fibres/mm². Density of large myelinated fibres was 1897±652 fibres/mm² (61% of control), whereas that of small myelinated fibres was 2564±832 fibres/mm² (50% of control), indicating a relatively predominant reduction of small myelinated fibres in most cases. Remarkable oedema in the endoneurium was present in most cases (fig 1A). The total endoneurial area was significantly increased compared with controls (1.725±0.434 mm² vs 1.089±0.061 mm², p<0.01; fig 2A). The estimated total number of myelinated fibres in the complete cross-section of the sural nerve was significantly reduced compared with controls (6367±2123 fibres vs 8684±959 fibres, p<0.05; fig 2B). The number of large myelinated fibres was reduced, but not to a significant extent (2663±920 fibres vs 3289±494 fibres; fig 2B), whereas the number of small myelinated fibres was significantly reduced (3704±1340 fibres vs 5396±628 fibres, p<0.05; fig 2B). Axonal sprouting of myelinated fibres and onion-bulb formation were not conspicuous in any case compared with controls.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Sural nerve biopsy specimen from a patient with polyneuropathy, organomegaly, endocrinopathy, monoclonal gammopathy and skin changes (POEMS) syndrome. (A) In light microscopy, subperineurial oedema is conspicuous. (B, C and D) In electron microscopy, uncompacted myelin lamellae are more frequent in small myelinated fibres. (E) Unmyelinated fibres are preserved, whereas a remnant of myelin sheath, suggestive of axonal degeneration of myelinated fibres, is seen. (F) Teased fibres from a patient with POEMS syndrome. Segmental demyelination resulted from widening of nodes of Ranvier (arrows) is conspicuous in small myelinated fibres. Scale bars = 50 μm (A), 1 μm (B and C) and 3 μm (D and E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Morphometric assessments of sural nerve biopsy specimens. Black columns indicate polyneuropathy, organomegaly, endocrinopathy, monoclonal gammopathy and skin changes (POEMS) syndrome and white columns indicate controls. Each error bar represents the standard deviation. (A) Total endoneurial area. Total endoneurial area was markedly enlarged compared with controls (p<0.01). (B) Estimated number of myelinated fibres per complete cross-section of the sural nerve. The estimated number of total myelinated fibres per sural nerve cross-section was significantly reduced compared with controls (p<0.05). There was no significant difference in the number of large myelinated fibres between POEMS syndrome and controls, but there was a significant difference in small myelinated fibres (p<0.05). (C) The estimated number of unmyelinated fibres per complete cross-section of the sural nerve. In contrast to the reduction of myelinated fibres, there was no difference in the number of unmyelinated fibres. (D) Percentage of subunits with unmyelinated fibres in the total population of Schwann cell subunits associated with unmyelinated fibres. The percentage was not different from that of the control group, suggesting that unmyelinated fibres were preserved. (E) The percentage of fibres with uncompacted myelin lamellae in myelinated fibres. The percentage was significantly more frequent in small myelinated fibres than in large myelinated fibres (p<0.001). (F) Teased-fibre studies. The frequency of fibres with segmental demyelination due to widened nodes of Ranvier was significantly more frequent in small fibres than in large fibres (p<0.0001). *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001 with Mann–Whitney U test. MF, myelinated fibres; UMF, unmyelinated fibres.

In contrast to myelinated fibres, a reduction of unmyelinated fibres was not apparent. Although the densities of unmyelinated fibres ranged from 15736 to 32323 fibres/mm² and significantly decreased compared with controls (22947±5032 fibres/mm² vs 30655±2731 fibres/mm², p<0.01), the estimated total number of unmyelinated fibres in the complete cross-section of the sural nerve in POEMS syndrome was almost the same as that seen in controls (32932±6570 fibres vs 32348±4174 fibres; fig 2C). The population of unmyelinated fibres was preserved, even in cases with a marked loss of myelinated fibres. In two patients who had impaired glucose tolerance (patients 13 and 16), ballooning of unmyelinated fibres, suggestive of degenerating fibres, and clusters of small unmyelinated fibres, suggestive of regenerating fibres, were observed. One of the patients (patient 13) had increased numbers of unmyelinated fibres due to the regenerating fibres. Findings suggestive of unmyelinated fibre degeneration or regeneration were not obvious in other cases. The percentage of Schwann cell subunits with unmyelinated fibres was not different from that in controls (78.3±9.5% vs 79.6±8.3%; fig 2D). This finding suggests that there was no increase of empty subunits and further supports the view that unmyelinated fibres were preserved.21

The percentage of fibres with UML was 1.7±1.7%. They were more frequent in small myelinated fibres than in large myelinated fibres (2.5±2.4% vs 0.5±1.0%, p<0.001; fig 1B, C, D, and 2E).

In teased-fibre studies, irregularity of myelin was conspicuous and both segmental demyelination and axonal degeneration were frequently found. Segmental demyelination resulted from widening of the nodes of Ranvier, and was more frequently found in small fibres (41.8±17.6% of small fibres vs 14.2±19.3% of large fibres, p<0.0001; fig 1C, 2F).

The total myelinated fibres per complete cross-section of the sural nerve was significantly less in the group with hyperalgesia than those without it (5385±1895 fibres vs 8051±1304 fibres, p<0.01; fig 3A). Both large and small myelinated fibres were significantly reduced in the group with hyperalgesia compared with that without hyperalgesia (2290±820 fibres vs 3304±741 fibres, p<0.05 for large myelinated fibres; 3095±1144 fibres vs 4747±988 fibres, p<0.01 for small myelinated fibres; fig 3A). Compared with controls, the number of both large and small myelinated fibres was significantly reduced in the group with hyperalgesia (p<0.05 for large myelinated fibres, p<0.01 for small myelinated fibres). On the other hand, in the group without hyperalgesia, the number of large myelinated fibres was not reduced compared with controls, and the number of small myelinated fibres was only slightly reduced. There was no difference in the number of unmyelinated fibres in the complete cross-section of the sural nerve between the groups with and without hyperalgesia (32308±6010 fibres vs 33912±7764 fibres; fig 3B). As for the frequency of fibres with UML and the frequency of segmental demyelination in teased-fibre studies, these were not significantly different between the groups with and without hyperalgesia except for a higher frequency of segmental demyelination in small fibres in the group with hyperalgesia than that without it (p<0.01).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 The relationship between hyperalgesia and total number of nerve fibres per complete cross-section of the sural nerve. Black columns indicate the group with hyperalgesia and white columns indicate the group without hyperalgesia. Each error bar represents the standard deviation. (A) The reduction of total myelinated fibres was significant in the group with hyperalgesia compared to those without it (p<0.01). Both large and small myelinated fibres were significantly reduced (p<0.05 for large myelinated fibres, p<0.01 for small myelinated fibres). (B) No difference in the number of unmyelinated fibres was observed between the groups with and without hyperalgesia. *p<0.05 and **p<0.01 with Mann–Whitney U test. MF, myelinated fibres; UMF, unmyelinated fibres.

Cytokine profiles and their correlation to neuropathic pain

Individual cases showed extensive variation of the concentration for proinflammatory cytokines (IL-1β, IL-6 and TNF-α) in the group with POEMS syndrome (fig 4A, B and C). As a whole, a significant increase was found only for IL-6 (p<0.01). Levels of these proinflammatory cytokines for the POEMS syndrome and control groups were 17.3±25.9 and 5.6±5.4 pg/ml for IL-1β, 45.4±96.2 and 2.3±2.5 pg/ml for IL-6, and 138.7±349.5 and 2.8±6.7 pg/m for TNF-α. Levels of the anti-inflammatory cytokine (IL-1ra) were significantly lower in the POEMS syndrome group than in the control group (175.3±71.7 and 469.9±146.2 pg/ml, p<0.001; fig 4D). Levels of cytokine receptors (sIL-6r, sTNFrI and sTNFrII) were significantly higher in the group with POEMS syndrome than the control group (fig 4E, F and G). These levels for the POEMS syndrome and control groups were 28.9±6.6 and 22.6±4.1 ng/ml (p<0.05) for sIL-6r, 2241.7±671.2 and 1241.7±175.3 pg/ml (p<0.001) for sTNFrI, and 3178.3±738.1 and 2223.3±401.7 pg/ml (p<0.01) for sTNFrII.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4 Serum cytokine and cytokine receptor levels in polyneuropathy, organomegaly, endocrinopathy, monoclonal gammopathy and skin changes (POEMS) syndrome and control groups. Filled circles represent POEMS syndrome patients with hyperalgesia, open circles represent POEMS syndrome patients without hyperalgesia, and open squares represent controls. (A) Interleukin-1β (IL-1β). (B) Interleukin-6 (IL-6). (C) Tumour necrosis factor-α (TNF-α). (D) IL-1 receptor antagonist (IL-1ra). (E) Soluble IL-6 receptor (sIL-6r). (F) Soluble TNF receptor I (sTNFrI). (G) Soluble TNF receptor II (sTNFrII). All patients who showed extensive elevation of proinflammatory cytokines (IL-1β, IL-6 and TNF-α) had hyperalgesia. All of these proinflammatory cytokines were significantly elevated in the group with hyperalgesia than that without it (p<0.05 for IL-1β and IL-6, p<0.01 for TNF-α). On the other hand, no significant difference between the two groups was found for IL-1ra, sIL-6r, sTNFrI and sTNFrII. Compared with controls, the group with hyperalgesia showed significant elevation for IL-1β (p<0.05), IL-6 (p<0.01), TNF-α (p<0.05), sTNFrI (p<0.01) and sTNFrII (p<0.05), and reduction for IL-1ra (p<0.01). The group without hyperalgesia showed significant elevation for sTNFrI (p<0.01) and sTNFrII (p<0.05) and reduction for IL-1ra (p<0.01) compared with controls. *p<0.05 and **p<0.01 with Mann–Whitney U test. NS, not significant.

All patients who showed extensive elevation of proinflammatory cytokines (IL-1β, IL-6 and TNF-α) had hyperalgesia. All of these proinflammatory cytokines were significantly elevated in the group with hyperalgesia compared to the group without it (p<0.05 for IL-1β and IL-6, p<0.01 for TNF-α; fig 4A, B and C). No significant difference between the two groups was found for IL-1ra, sIL-6r, sTNFrI and sTNFrII (fig 4D, E, F and G). Compared with controls, the group with hyperalgesia showed significant elevation of IL-1β (p<0.05; fig 4A), IL-6 (p<0.01; fig 4B), TNF-α (p<0.05; fig 4C), sTNFrI (p<0.01; fig 4F) and sTNFrII (p<0.05; fig 4G), and reduction of IL-1ra (p<0.01; fig 4D). The group without hyperalgesia showed significant elevation for sTNFrI (p<0.01; fig 4F) and sTNFrII (p<0.05; fig 4G) and reduction for IL-1ra (p<0.01; fig 4D) compared with controls. In patient 7, who complained of severe spontaneous pain without hyperalgesia, none of the proinflammatory cytokines was elevated compared with controls. On the other hand, proinflammatory cytokines were examined in two patients (patients 1 and 13) who had hyperalgesia without spontaneous pain and all cytokines were elevated in both patients.

DISCUSSION

Because POEMS syndrome affects a wide range of organs, its detailed neuropathic features have not yet been fully described. Evaluating the extent of nerve fibre loss in sural nerve biopsy specimens from POEMS syndrome patients is difficult due to marked enlargement of the endoneurial area due to extensive oedema. When nerve fibre loss is estimated by density, the extent may be overestimated when the cross-sectional area is enlarged. To exclude this possibility, we estimated the total number of nerve fibres in the compelete cross-section of the sural nerve and determined the precise extent of nerve fibre loss. Thus, we were able to clearly demonstrate the clinical and pathological correlates of POEMS syndrome. Although sensory symptoms were variable, ranging from the positive to the negative, the most characteristic feature in our series was the presence of pain. More than half of our patients reported uncomfortably painful symptoms, including spontaneous pain and hyperalgesia. Indeed, many of the patients were referred to the hospital due to painful symptoms. On the other hand, patients with CIDP are predominantly characterised by motor weakness rather than sensory complaints, although painful symptoms are reported.25 Both POEMS syndrome and CIDP are associated with demyelination in the peripheral nervous system, and electrophysiological findings are, therefore, to some extent similar. Indeed, some of our patients were initially diagnosed with CIDP before consultation to our hospital for sural nerve biopsy, particularly when the associated symptoms other than neuropathy were not conspicuous. The rarity of this syndrome also makes it difficult to diagnose correctly. Recognition of the characteristic clinical features, including spontaneous pain and hyperalgesia, may be useful in discriminating patients with POEMS syndrome from those with CIDP.

The mechanism as to why painful symptoms occur in a subgroup of patients with POEMS syndrome needs to be clarified. It is interesting that neuropathy of equal aetiology can be painful or painless. The prototype of known painful neuropathies, such as familial amyloid polyneuropathy, alcoholic neuropathy, Fabry disease, and the subgroups of polyneuropathies associated with diabetes mellitus, paraneoplastic syndrome and Sjögren’s syndrome, is characterised by small-fibre predominant axonal degeneration with relative preservation of large myelinated fibres.16 18 20 26^–29 In these neuropathies, unmyelinated fibres are the most profoundly affected and demyelinating change is not a primary feature. On the other hand, our POEMS syndrome patients revealed extensive demyelination and well-preserved unmyelinated fibres. Although we need to further assess whether the most distal portion of unmyelinated fibres are affected at the epidermal site, the mode of nerve fibre injury in the sural nerve biopsy specimens in POEMS syndrome was distinctive from that of known painful neuropathies with predominant small-fibre loss. According to a previous study,30 hyperalgesia comprises a dynamic component (brush-evoked pain, allodynia) that is signalled by large myelinated afferents and a static component (hyperalgesia to pressure stimuli) that is signalled by unmyelinated afferents. Hyperalgesia in our series was similar to the latter.

Although the mechanism of neuropathic pain has been intensively investigated, existing knowledge has been based mainly on animal research or experimental studies of healthy human subjects. As for the pathological condition of human neuropathic pain, post-herpetic neuralgia, complex regional pain syndromes and diabetic neuropathy have been relatively well-investigated,31^–33 but little is known about neuropathic pain with other aetiologies. Neuropathic pain has been classified and studied according to its nature rather than to the nosology of the disease. Hyperalgesia has been thought to conduct through afferent A-fibres,34 but recent observations suggest that hyperalgesia is also related to afferent C-fibres.33^–35 In our case, myelinated fibres were more profoundly affected in patients with hyperalgesia than those without it. In contrast, unmyelinated fibres were not reduced in both groups. These observations suggest that myelinated fibre injury, rather than unmyelinated fibre injury, is related to the appearance of painful symptoms in POEMS syndrome. This is similar to models of cold hyperalgesia in healthy human subjects, which suggest that hyperalgesia is induced by decreased inhibition of activated C-fibres due to the blockade of A-delta fibres.36 In post-herpetic neuralgia, the reduction in skin unmyelinated fibre innervation was inversely correlated with severity of allodynia, suggesting that the presence of surviving unmyelinated fibres is important for the induction of allodynia.31 Thus, neuropathic pain can be induced most effectively when unmyelinated C-fibres are well-preserved in the population under the condition that inhibition of C-fibre activity by myelinated A-delta fibres is lacking. The existence of well-preserved C-fibres and decreased A-fibres, including A-delta fibres, in POEMS syndrome is similar to the conditions in which neuropathic pain is effectively provoked. Taken together, the painful symptoms in POEMS syndrome could be generated by well-preserved afferent C-fibres when the inhibitory effect of afferent A-fibres is reduced.

In addition, serum levels of proinflammatory cytokines, including IL-1β, IL-6 and TNF-α, are known to be increased in patients with POEMS syndrome, although some variation exists.3 37 38 Our data also show extensive variation in the levels of these proinflammatory cytokines among individual patients; they are not necessarily elevated. However, the elevation of these cytokines seemed to be related to the presence of hyperalgesia in our cases. The source of these cytokines in patients with POEMS syndrome remains uncertain as they are produced by a variety of host cells and tumour cells. For example, proinflammatory cytokines are known to be produced from Schwann cells undergoing axonal degeneration.39 On the other hand, the cytokine itself may induce axonal degeneration.40 In our cases, levels of one of the proinflammatory cytokines, IL-6, is positively correlated to the degree of myelinated fibre loss. It is interesting that these proinflammatory cytokines are known to be closely related to provocation of neuropathic pain.41

In summary, the painful symptoms in POEMS syndrome may be generated through well-preserved afferent C-fibres when the inhibition of C-fibres by A-fibres is decreased, especially in the presence of cytokine sensitisation, thus providing new insight into the pathophysiology of neuropathic pain.