Article Text


Research paper
Proposal of new clinical diagnostic criteria for POEMS syndrome
  1. Tomoki Suichi1,
  2. Sonoko Misawa1,
  3. Yasunori Sato2,
  4. Minako Beppu1,
  5. Emiko Sakaida3,
  6. Yukari Sekiguchi1,
  7. Kazumoto Shibuya1,
  8. Keisuke Watanabe1,
  9. Hiroshi Amino1,
  10. Satoshi Kuwabara1
  1. 1 Department of Neurology, Graduate School of Medicine, Chiba University, Chiba, Japan
  2. 2 Department of Global Clinical Research, Graduate School of Medicine, Chiba University, Chiba, Japan
  3. 3 Department of Hematology, Graduate School of Medicine, Chiba University, Chiba, Japan
  1. Correspondence to Dr Satoshi Kuwabara, Department of Neurology, Graduate School of Medicine, Chiba University, Chiba 260-8670, Japan; kuwabara-s{at}


Objective To propose the optimal diagnostic criteria for polyneuropathy, organomegaly, endocrinopathy, M-protein and skin changes (POEMS) syndrome using appropriate statistical methods and disease controls.

Methods This retrospective cohort study included 104 consecutive patients with suspected POEMS syndrome, among whom a gold standard group of 60 patients with definitive POEMS syndrome diagnosis were followed for at least 12 months to strictly exclude other disorders mimicking POEMS syndrome and to confirm response to POEMS syndrome-specific treatment. Thirty patients with chronic inflammatory demyelinating polyradiculoneuropathy (demyelinating polyradiculoneuropathy controls) and 30 with multiple myeloma or immunoglobulin light chain amyloidosis (monoclonal plasma cell proliferation controls) were also included. Logistic regression analyses were performed to determine optimal combination of clinical and laboratory abnormalities, characteristic of POEMS syndrome.

Results The diagnostic criteria were statistically defined as the presence of the three major criteria (polyneuropathy (typically demyelinating), monoclonal plasma cell proliferative disorder and elevated vascular endothelial growth factor) and at least two of the four minor criteria (oedema/effusion, skin changes, organomegaly and sclerotic bone lesions), based on best performance by area under the receiver operating characteristic curve analyses. The sensitivity and specificity were 100% and 100%, respectively; the diagnostic accuracy of the proposed criteria was equivalent to somewhat complicated previous criteria.

Conclusions The statistically defined, simple diagnostic criteria for POEMS syndrome could accelerate early diagnosis and treatment, thereby contribute to better outcome in patients with this serious disease. Prospective larger studies are required to confirm the validity.

  • POEMS syndrome
  • diagnostic criteria
  • sensitivity
  • specificity

Statistics from


Polyneuropathy, organomegaly, endocrinopathy, M-protein and skin changes (POEMS) syndrome is a rare cause of demyelinating polyneuropathy associated with plasma cell dyscrasia and vascular endothelial growth factor (VEGF) overproduction.1 2 In addition to the five listed features, there are several other important manifestations characteristic of this disorder, such as oedema/effusion, sclerotic bone lesions, papilloedema and thrombocytosis.3–5

Diagnostic criteria for POEMS syndrome that have been published to date are based on a combination of characteristic clinical and laboratory findings.3–7 However, these criteria were not validated with appropriate disease controls associated with demyelinating neuropathy or M-protein; therefore, there is no very firm evidence of their diagnostic accuracy. Additionally, these currently used criteria include too many parameters, which can lead to complications during the diagnosis of these patients by neurologists, haematologists or general practitioners.

POEMS syndrome used to be a devastating disease, with a reported median survival time of 33 months in the 1980s,8 when patients were treated primarily merely with corticosteroids. Since around the 2000, therapeutic interventions available for multiple myeloma, such as autologous peripheral blood stem cell transplantation, thalidomide/lenalidomide and proteasome inhibitor therapies, have been utilised in patients with POEMS syndrome, which have improved its prognosis substantially.5 6 However, many patients do not improve sufficiently largely because of delayed diagnosis, which results in poor outcomes.7 Diagnosis of POEMS syndrome in the early disease stage is often challenging, particularly in patients who have polyneuropathy as an isolated symptom. In fact, patients with POEMS syndrome are often misdiagnosed with chronic inflammatory demyelinating polyradiculoneuropathy (CIDP).9 10

The aim of this study was to propose optimal and simple diagnostic criteria, which are essential for early diagnosis and treatment of POEMS syndrome with high sensitivity and specificity using logistic regression analyses.


Study design

In this retrospective cohort study, we defined a gold standard group of patients with POEMS syndrome and reviewed their medical records to explore the frequencies of characteristic signs and symptoms. Patients with CIDP (demyelinating polyradiculoneuropathy controls) and those with multiple myeloma, immunoglobulin light chain amyloidosis (AL amyloidosis) or monoclonal gammopathy of undetermined significance (MGUS) as monoclonal plasma cell proliferation controls were also included to clarify the specificity of the current criteria for POEMS syndrome.


A total of 104 consecutive patients with suspected POEMS syndrome who were seen at Chiba University Hospital between 2000 and 2015 were screened (figure 1). Of these, 12 patients with a different final diagnosis,16 patients who were previously treated with irradiation, alkylators, immunomodulatory drugs, proteasome inhibitors, anti-VEGF monoclonal antibodies or autologous peripheral blood stem cell transplantation before referral and six patients without detectable monoclonal plasma cell proliferation were excluded and 70 treatment-naive patients were diagnosed with POEMS syndrome. Next, patients exhibiting typical clinical features and treatment responses consistent with POEMS were followed up at our hospital for at least 1 year after treatment. The patient group (n=60) was defined as a gold standard cohort of POEMS syndrome (figure 1).

Figure 1

Flow diagram of assessed patients. Between 2000 and 2015, 104 patients with suspected POEMS syndrome were screened. Among them, 12 patients who received other diagnoses, 16 patients who were initiated on treatment (thalidomide, lenalidomide, bortezomib, transplantation, radiation or chemotherapy) and six patients without detectable plasma cell proliferation were excluded. The remaining 70 patients were initiated treatment and follow-up for 1 year. During the follow-up, six patients died and four patients were lost to follow-up. Therefore, 60 patients were defined as gold standard POEMS syndrome group. In addition, 30 demyelinating polyradiculoneuropathy control patients and 30 monoclonal plasma cell proliferation control patients were included. AL, immunoglobulin light chain amyloidosis; CIDP, chronic inflammatory demyelinating polyradiculoneuropathy; MGUS, monoclonal gammopathy of undetermined significance; MM, multiple myeloma; POEMS, polyneuropathy, organomegaly, endocrinopathy, M-protein and skin changes.

We also studied 30 patients with CIDP as the demyelinating polyradiculoneuropathy controls and 30 with multiple myeloma (n=27), AL amyloidosis (n=1), MGUS (n=2) as monoclonal plasma cell proliferation controls. Typically, CIDP was diagnosed on the basis of the European Federation of Neurological Societies/Peripheral Nerve Society criteria.11 Multiple myeloma and MGUS diagnoses were based on the criteria from international myeloma working group 2003.12 AL amyloidosis was diagnosed on the basis of the consensus criteria.13


The medical records of patients were reviewed to determine the frequencies of signs/symptoms and laboratory abnormalities characteristic to POEMS syndrome. Assessments of each manifestation was performed as follows: (1) polyneuropathy: distal-dominant sensory and motor symptoms, hyporeflexia or areflexia and evidence of demyelination confirmed by systematic nerve conduction studies; (2) monoclonal plasma cell proliferative disorder: M-protein detected by serum immunofixation or monoclonal plasma cell proliferation revealed by bone mallow aspiration/biopsy. Identification of plasma cell clonality was assessed by immunohistostaining, in situ hybridisation and/or flow cytometry ; (3) serum VEGF elevation: >1000 pg/mL measured by ELISA (Special Reference Laboratory, Tokyo, Japan)14; (4) extravascular volume overload: oedema or effusion detected by chest and abdominal CT scan; (5) skin changes: evident hyperpigmentation, hypertrichosis, glomeruloid haemangiomata, white nails or cyanosis; (6) organomegaly: hepatosplenomegaly or lymphadenopathy demonstrated by CT scan or ultrasonography; (7) sclerotic bone lesions: bone sclerosis detected by X-ray or CT, bone or positron emission tomography scans; (8) endocrinopathy: consistent clinical features and each hormonal abnormality; (9) papilloedema: evaluation by an ophthalmologists or neurologists; (10) thrombocytosis: blood platelet count >350×109/L and (11) Castleman disease: characteristic clinical and laboratory features and lymph node biopsy findings.

Standard protocol approvals, registrations and patient consents

Because this study is conducted by retrospective chart reviews and have no risk for patients, written informed consents from the patients were waived. Information explaining the study and option to refuse participation was announced on the online homepage of the Department of Neurology, Chiba University Graduate School of Medicine.

Statistical analysis

For baseline variables, summary statistics were constructed employing frequencies and proportions. First, Spearman’s rank-order correlation coefficients between variables were calculated, and for group containing two variables where the correlation coefficient was 0.9 or greater, one representative variable was chosen and the other variable was discarded. Next, best combination of variables was selected using multivariate logistic regression model with a stepwise selection procedure. Specifically, the stepwise procedure was set with a cut-off at a priori-defined p values of<0.05 for inclusion and >0.05 for exclusion. Additionally, Akaike information criterion (AIC) was applied to determine the best model among those tested.15 At each step, one variable was deleted, and the model with the smallest AIC value was selected, until any further deletion caused an increase in AIC value. The model’s calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test, and the predictive accuracy was assessed on basis of area under the receiver operating characteristic (ROC) curve (AUC). AUC was estimated using a form of the trapezoid method, and the 95% CI of the AUC was estimated by the bootstrap method.16 The sensitivity and specificity of the newly proposed criteria and the previously published criteria were evaluated.3–6

All statistical analyses were performed by SAS software V.9.4 by one of the authors (YS), a biostatistician, who was blind to clinical information.


Clinical laboratory characteristics

A total of 60 patients were categorised as the gold standard group of patients with POEMS syndrome. Patients with typical CIDP (n=30) and those with multiple myeloma/AL amyloidosis/MGUS (n=30) were also included (figure 1).

Frequencies of the clinical manifestations and laboratory abnormalities of POEMS syndrome are shown in table 1. Among the gold standard POEMS group, all patients had demyelinating polyneuropathy, monoclonal plasma cell proliferative disorder and VEGF elevation. Extravascular volume overload and skin changes were also high prevalence among the patients with POEMS syndrome. Bone marrow biopsy and aspiration were performed in 48 (80%) and 57 (95%) patients, respectively. In nerve conduction studies, median or ulnar motor nerve conduction velocity was reduced in 93%, and minimal F-wave latency was prolonged in 97% of the patients.

Table 1

Frequencies of clinical manifestations and laboratory abnormalities

Among the CIDP group, monoclonal plasma cell proliferative disorder and elevation of VEGF levels were present in five (17%) and four (13%) patients, respectively, whereas other systemic manifestations were very rare. The type of M-protein was as follows; two patients presented immunoglobulin (Ig)G kappa, and the remaining three patients presented IgG lambda, IgM lambda and biclonal (IgG lambda and IgM kappa), respectively.

Among the monoclonal plasma cell proliferation control group, axonal polyneuropathy and serum VEGF elevation were found in nine (30%) and one (3%) patients, respectively, whereas oedema/effusion was found in 10 (33%) patients; however, other systemic symptoms were rare. The type of M-protein was as follows: IgG kappa in nine, IgG lambda in four, IgA kappa in two, IgA lambda in four, Bence-Jones protein kappa in five, Bence-Jones protein lambda in three and biclonal (IgG kappa and IgA kappa) in one, respectively. Data of two patients are not clearly classified.

Logistic regression analysis of diagnostic procedures

In the initial step, we included the three major criteria (polyneuropathy, monoclonal proliferative disorder and VEGF elevation) given that they are core features of POEMS syndrome that have been traditionally employed in currently used diagnostic criteria and were the most frequently observed criteria in the present study. Additionally, logistic regression and ROC analysis identified four minor criteria: extravascular volume overload, skin changes, organomegaly and sclerotic bone lesions, based on the best performance by AUC (online supplementary file 1) and AIC values.

Supplementary data

Diagnostic criteria for POEMS syndrome were therefore defined as presence of all of the three major criteria and at least two of the four minor criteria (table 2).

Table 2

Proposed and existing diagnostic criteria of POEMS syndrome

Comparison of the proposed criteria with the existing criteria

Four parameters from the existing criteria were excluded in the criteria analysed in the present study (table 2). The sensitivity and specificity of the proposed criteria were 100% and 100%, respectively, which were equivalent to those of the existing criteria (table 3).

Table 3

Comparison of sensitivity/specificity between the diagnostic criteria


In this study, we modified the currently used diagnostic criteria for POEMS syndrome based on a statistically optimal combination of clinical manifestations in patients with definitive POEMS syndrome. We first demonstrated the sensitivity and specificity by utilising disease controls such as demyelinating polyneuropathy and monoclonal plasma cell proliferation controls. In the proposed criteria, several parameters are removed from existing criteria. We successfully simplified diagnostic criteria without deteriorating sensitivity and specificity.

In the present study, all patients with POMES syndrome had polyneuropathy, monoclonal plasma cell proliferative disorder and VEGF elevation, which are core features of POEMS syndrome. We admitted these features as mandatory criteria according to statistical analysis; however, detection of monoclonal plasma cell proliferation and polyneuropathy should be done carefully. A nationwide survey conducted in Japan reported that among 102 patients with POEMS syndrome, 25% lacked M-protein.8 The study dated back to 1980s, and most patients included were examined by immunoelectrophoresis. Its low sensitivity was responsible for the lack of M-protein, whereas our patients were examined by immunofixation, which resulted in percentage of the lack of M-protein dropped to 2%. If M-protein is negative on immunofixation, bone marrow biopsy, positron emission tomography (PET), free light chain and Bence-Joes protein should be tested to search monoclonal plasma cell proliferation. Regarding polyneuropathy, the case of patients with POEMS syndrome with no evidence of polyneuropathy was reported.17 We consider that the diagnosis of POEMS syndrome should be judged as a whole core abnormalities especially in such atypical case. In addition, Castleman disease variant of POEMS syndrome should be considered separately.5

The currently utilised criteria for the diagnosis of POEMS syndrome, which includes a total of 11 items including two mandatory major criteria, three other major criteria and six minor criteria,4–6 are somewhat complicated. Therefore, we identified four minor criteria based on best performance by statistical analyses. Altogether, these revisions do not deteriorate the sensitivity/specificity of POEMS syndrome diagnosis.

One other unique aspect of the present study was disease control groups, demyelinating polyradiculoneuropathy controls including patients with CIDP and monoclonal cell proliferation controls including multiple myeloma, AL amyloidosis and patients with MGUS. Some patients with CIDP have monoclonal gammopathy,18 whereas M-protein patients can also develop polyneuropathy.19–21 These patients are potentially misdiagnosed with POEMS syndrome; in fact, 60% of patients with neuropathy-onset POEMS syndrome were initially diagnosed as CIDP.10 As evidenced by high specificity revealed by our analysis, our new criteria can exclude these disorders that can mimic POEMS syndrome.

Additionally, it is actually difficult to exclude patients with combination of multiple diseases (eg, idiopathic axonal neuropathy, MGUS, elevated serum VEGF from iron deficiency/anaemia/hypoxia, oedema/effusion from heart failure and acrocyanosis from arteriosclerosis). A good history and physical examination followed by appropriate testing, particularly nerve conduction studies, can differentiate this syndrome from other conditions. In nerve conduction studies of POEMS syndrome, median motor nerve conduction velocity is reduced and sever axonal loss is found in lower extremity.9 In any case, making the diagnosis of POEMS syndrome can originally be a challenge and careful exclusion of mimicking diseases should be done before potentially high-risk therapy such as autotransplantation.

There are several limitations in the present study. First, this was a retrospective study. Patient screening was not standardised, and several data were not available; particularly, assessments of bone marrow, sclerotic bone lesions and papilloedema were insufficient. Bone marrow was not evaluated in three patients with POEMS syndrome and 30 patients with CIDP. Second, this study was conducted at a single neurology centre, and selection bias could not be excluded; most of the patients were referred because of polyneuropathy. Furthermore, lack of validation cohort is also a limitation. Prospective multicenter studies involving both neurology and haematology departments are necessary to validate the utility of the proposed criteria. However, there are an estimated 340 patients with POEMS syndrome in Japan,22 and the present study screened 104 patients (one-third of Japanese patients with POEMS); therefore, we believe that the findings of the present study can be generalised for proposal of new diagnostic criteria for POEMS syndrome. Third, the numbers of demyelinating polyradiculoneuropathy controls (n=30) and monoclonal plasma cell proliferation controls (n=30) were insufficient. This limitation might affect the specificity of proposed criteria and a validation study will be required. We also have to raise false-positive conditions potentially associated with elevated serum VEGF, including iron deficiency,23 anaemia,24 hypoxia25 and various types of cancer.25 Patients with breast, gastrointestinal or prostate cancer present further increase in VEGF levels on metastatic stage.25 These conditions along with unrelated neuropathy, M-protein and minor criteria possibly fulfil proposed criteria.

In conclusion, we proposed new diagnostic criteria for POEMS syndrome; the criteria developed on the basis of statistical validity and exhibit high sensitivity and specificity. The proposed criteria should enable early diagnosis and appropriate treatments, thereby contributing to better outcome in patients with this potentially devastating disorder.


View Abstract


  • Contributors SM, TS and SK conceived the study concept and design and drafted the manuscript; MB, ES, YS, KS, KW and HA enrolled patients and collected the data and YS performed statistical analyses. All authors approved the final manuscript for submission.

  • Funding This work was supported in part by the Health and Labour Sciences Research Grant on Intractable Diseases (Neuroimmunological Diseases) and the Research Grant 16B-1 for Nervous and Mental Disorders from the Ministry of Health, Labour and Welfare of Japan.

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval The ethics committee of Chiba University Graduate School of Medicine.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.