Nonsystemic Vasculitic Neuropathy: Insights from a Clinical Cohort

M.P. Collins, M.D.

Neurosciences Department

Marshfield Clinic

Marshfield, Wisconsin

M.I. Periquet, M.D.

Department of Neurology

The OhioStateUniversityCollege of Medicine

Columbus, Ohio

J.R. Mendell, M.D.

Department of Neurology

The OhioStateUniversityCollege of Medicine

Columbus, Ohio

Z. Sahenk, M.D., Ph.D.

Department of Neurology

The OhioStateUniversityCollege of Medicine

Columbus, Ohio

H.N. Nagaraja, Ph.D.

Department of Statistics

The OhioStateUniversity

Columbus, Ohio

J.T. Kissel., M.D.

Department of Neurology

The OhioStateUniversityCollege of Medicine

Columbus, Ohio

Supplementary Content

Disclosures: None

Character count (title): 66Word count (abstract): 260Word count (text): 4751

Corresponding Author: Michael P. Collins, M.D.

Neurosciences Department

The Marshfield Clinic

1000 N. Oak Avenue

Marshfield, WI54449


Fax: (715)-387-5727



Background: Nonsystemic vasculitic neuropathy (NSVN) is an uncommon disorder. Few series with small numbers of patients have been reported. The prognosis and treatment of patients presentingwith NSVN remain uninvestigated. We sought to address these issues by assembling a large retrospective cohort with extended follow-up. Methods: All nerve biopsies performed over 20 years were reviewed; cases with definite, probable, or possible vasculitis were segregated for clinical correlation. Patients satisfying clinical criteria for NSVN at presentation were selected. Clinicopathologic, treatment, and outcome measures were analyzed in patients followed for  6 months. Results: 48 patients (30 women, 18 men) with median 63 months follow-up were identified. Most patients (85%) had extensive, overlapping involvement of multiple nerves. Only one had a symmetric polyneuropathy. Most neuropathies (96%) were painful. In 96%, nerve damage was distally-accentuated, but most had concurrent proximal weakness. Diagnostic sensitivity was 58% for superficial peroneal nerve/peroneus brevis muscle biopsy and 47% for sural nerve biopsy. Combination corticosteroid/cytotoxic therapy was more effective than corticosteroid monotherapy in inducing remission and improving disability, with trends toward reduced relapses and chronic pain. Treatment with cyclophosphamide for > 6 months decreased the relapse rate, which was 46% for all patients. Disease/treatment-related mortality was 10%. Six percent developed cutaneous involvement. Although chronic pain persisted in 60% of survivors, 80% had good outcomes. Conclusions: NSVN nearly always presents as an asymmetric, distally-accentuated, painful, sensorimotor polyneuropathy. Risks for systemic spread and death are small, and, aside from pain, neurologic prognosis is unexpectedly good. Although not a randomized controlled trial, combination therapy produced the best outcome in this cohort.


In vasculitis, blood vessels become inflamed and partially destroyed, thereby damaging tissues through ischemic, inflammatory, and immune mechanisms.1 Most classifications of vasculitis include only systemic disorders that affect multiple organs.2,3 However, vasculitis can also present in a “nonsystemic” fashion, confining itself to a single organ, such as skin,4 gastrointestinal5 or reproductive tract,6 lungs,7 kidneys,8 or CNS.9 Vasculitis restricted to the peripheral nervous system (PNS) was first recognized by Kernohan and Woltman in 1935.10 The concept then lay dormant until 1985 when seven such patients were described.1 In 1987, Dyck and coworkers reported 20 patients and coined the term “nonsystemic vasculitic neuropathy (NSVN).”11 Small numbers of patients have since appeared.12-17 The largest cohort (25 patients) was reported in 1996.18 The two largest NSVN cohorts were assembled retrospectively, with patients selected for absence of extra-neurologic involvement during follow-up.11,18 This selection criterion preordained a relatively benign prognosis and precluded an assessment of risk for systemic spread in patients presenting with a nonsystemic-appearing vasculitic neuropathy.

An important but overlooked question concerns appropriate treatment of patients with vasculitic neuropathy in general and NSVN in particular. There are no controlled data relevant to the treatment of any type of vasculitic neuropathy, apart from one study of Churg Strauss syndrome (CSS)-associated neuropathy.19 We have advocated combination therapy [corticosteroids (CS) and cyclophosphamide (CYC)] in NSVN, except for mild cases. Others favor CS monotherapy.11,20 These recommendations lack evidentiary support and are extrapolated from systemic vasculitis trials or anecdotal experience. This paucity of evidence prompted us to gather data on all NSVN patients treated at our institution over the past 20 years and analyze their long-term treatment outcomes. In contrast to other series, we included all patients meeting a predetermined definition of NSVN at the time of diagnosis, consistent with the manner in which clinicians actually function. Our goal was to systematically analyze treatment responses, clinical presentation, relapse rate, neurologic outcome, and mortality in patients with NSVN.


Pathologic selection criteria

All nerve biopsies (sural nerve, SPN/PBM, or superficial radial nerve) performed in the neuromuscular laboratory of The Ohio State University from 1981 to 2001 were reviewed. Biopsy procedures and histological techniques have been previously described.16,21 Biopsies were classified as showing (1) definite, (2) probable, or (3) possible vasculitis based on pre-established definitions.16Definite vasculitis was diagnosed if at least one blood vessel was infiltrated by inflammatory cells associated with signs of vascular injury, such as fibrinoid necrosis, endothelial cell disruption, fragmentation of the internal elastic lamina, hemorrhage, or acute thrombus. Probable vasculitis required transmural or perivascular inflammation unaccompanied by vascular destruction, but combined with at least one other supportive finding, including vascular thickening, luminal obliteration, recanalized thrombus, epineurial neovascularization, hemosiderin deposits, asymmetric nerve fiber loss, ongoing Wallerian-like degeneration, focal perineurial inflammation/thickening, or muscle fiber necrosis/regeneration. Possible vasculitis was diagnosed in non-inflammatory biopsies revealing predominant axonal alterations and one or more of four pathologic changes previously shown by our group to be predictive of PNS vasculitis: muscle fiber necrosis/regeneration, asymmetric nerve fiber loss, ongoing Wallerian-like degeneration, and vascular immune deposits.16

Patient selection criteria

For all patients meeting pathologic criteria for definite, probable, or possible vasculitis, available outpatient and inpatient records were reviewed. Each patient had been evaluated by at least one neuromuscular clinician from OhioState. In patients with pathologic diagnoses of probable or possible vasculitis, a subset with “clinically probable” vasculitic neuropathy was determined by application of published diagnostic criteria,16 consistent with precedent.11,18,22 All patients with definite or clinically probable vasculitic neuropathy were then divided into systemic versus nonsystemic groups according to standard classification schemes.2,3 Diagnosis of NSVN required (1) clinicopathologic evidence of vasculitic neuropathy; (2) no clinicopathologic signs of CNS or extra-neurologic involvement; and (3) no demonstrable etiology or predisposing condition (see Appendix). Patients were not excluded a priori for constitutional symptoms (weight loss, fever, arthralgias, anorexia, malaise, night sweats) or monoclonal gammopathy of undetermined significance (MGUS).

All patients satisfying criteria for NSVN were next assessed for adequacy of follow-up. For patients lacking at least two years follow-up at our institution, attempts were made to acquire additional information by (1) telephone interviews with patients and relatives, (2) retrieval of medical records from physician offices and hospitals, and (3) re-examination of selected patients. Patients who had less than six months follow-up after these efforts were excluded.

Data collection

In this manner, a cohort of 48 NSVN patients with  6 months follow-up was assembled. For each patient, historical, demographic, laboratory, and pathologic information was tabulated. Electrodiagnostic data was not tracked due to multiple examiners employing heterogeneous protocols. Charted treatment variables included agents, dosages, durations of therapy, adverse effects, and infectious complications (excluding upper respiratory syndromes, gastroenteritis, and oral/vaginal thrush). Treatment protocols were not standardized and at the discretion of the treating neurologist.

Outcome variables included: (1) pain (better, unchanged, worse); (2) sensory loss (better, unchanged, worse); (3) objective weakness (better, unchanged, worse); (4) composite Medical Research Council (MRC) strength score ( 4+ or 5- = 4.5, 4- or 3+ = 3.5, 3- = 2.5, and 2- = 1.5), obtained by summing the most-affected-side scores for shoulder abductors, elbow flexors, elbow extensors, wrist flexors, wrist extensors, hip flexors, knee flexors, knee extensors, ankle invertors, and ankle dorsiflexors; and (5) disability score using the Prineas scale.18,23,24 The primary outcome measure was “long-term response,” defined as positive if at least one of the preceding five outcome variables improved and none worsened, with improvement sustained for  6 months; and negative if any of the variables worsened or all were unchanged at 6 months. Patients who improved initially but worsened prior to 6 months were treated as non-responders (negative long-term response). In patients with a positive long-term response followed for  12 months, relapses were tracked. Relapse was defined as a new sign or symptom of weakness, sensory loss, or neuropathic pain after a sustained response to therapy. Secondary outcome measures for assessment of treatment responses were: (1) change in composite MRC score, (2) change in disability score, (3) relapse rate, (4) disease or initial treatment-related mortality rate, and (5) chronic pain in patients followed for at least 24 months. Final MRC and disability scores were determined at the most recent follow-up encounter, just prior to relapse, or just prior to death. All thromboembolic events occurring within six months of follow-up were categorized as disease/treatment-related, in view of the recognized potential of CS and pro-inflammatory cytokines (e.g., interleukin-1, tumor necrosis factor-) to promote thrombosis.25,26

Statistical Analyses

Statistical analyses were performed using SAS JMP, Version 4 (NC). For assessment of treatment responses, patients were divided into CS monotherapy and CS plus adjunct (combination therapy) groups according to their initialtreatment. The two groups were compared for potentially confounding variables such as age, neuropathic symptom duration, initial composite MRC score, initial disability score, erythrocyte sedimentation rate (ESR), and pathologic muscle involvement, using one-way analysis of variance or ANOVA (Student’s t test with equal variances and Welch ANOVA with unequal variances) for parametric analysis of continuous variables, Wilcoxan rank-sum test for nonparametric analyses, and Fisher’s exact test for comparison of qualitative variables. Differences between the two groups in long-term response rate were analyzed with Fisher’s exact test. Differences in relapse rate and chronic pain were assessed with the Pearson chi-square test. Comparisons between the groups for changes in composite MRC and disability scores were conducted with Wilcoxan rank-sum and t tests. The relationship between duration of CYC exposure and relapse rate was addressed with a 2 x 2 contingency table and Fisher’s exact test. Kaplan-Meier survival analysis was performed for the entire NSVN cohort, initial CS monotherapy subgroup, and initial CS plus adjunct subgroup. Survival data for the two subgroups was compared using the Mantel-Cox log-rank and Wilcoxan rank-sum tests. The influence of weight loss on long-term response rate and change in disability score for the entire NSVN cohort was evaluated with Fisher’s exact, Wilcoxan rank-sum, and t tests.

The initial CS group was next subdivided based on the presence or absence of an inaugural IV pulse of methylprednisolone. The effects of the IV pulse on long-term response rate, relapse rate, change in composite MRC score, and change in disability score were analyzed using similar techniques to the CS monotherapy versus combination therapy comparisons, with the addition of an analysis of covariance (ANCOVA) adjusting for the initial MRC and disability scores. To determine potential predictors of a long-term response to initial CS monotherapy, five clinical variables (age of onset, gender, symptom duration, initial composite MRC score, and initial disability scores) and four laboratory variables (ESR, antinuclear antibodies (ANA), rheumatoid factor (RF), and leukocytosis) were subjected to separate logistic regressions.

For assessment of complications, the incidence of CS-related side effects in patients exposed to prednisone was compared to the incidence of CYC-related side effects in patients exposed to CYC using Fisher’s exact test. Infection rates for patients treated with CS alone versus CS plus immunosuppressive agents were analyzed in a similar fashion. The yield of definite pathologic evidence of vasculitis in patients undergoing sural nerve compared to SPN/PBM biopsies was analyzed with Fisher’s exact test.


Clinical characteristics

Clinical characteristics of the cohort are summarized in Table E-1 on the Neurology Web site. Of the 48 NSVN patients, 14 were previously reported,1,16 albeit without the follow-up and treatment data obtained in this study. Thirty patients were women and 18 men (ratio 1.7:1). Age at onset of neuropathy ranged from 21 to 88 years (mean ± SD = 61.8 ± 14.5 years). Symptom duration prior to diagnosis varied from 2 weeks to 8 years (mean 8.9 ± 14.9 months; median 5 months). Weight loss (mean 21 ± 10.2 pounds) occurred in 35% and unexplained fevers in 15%.

Ninety-six percent of the neuropathies were painful, and most were sensorimotor in character. Only 13% of patients had predominantly or, less commonly, purely sensory findings. No patient presented with pure motor deficits. One patient had small fiber-restricted sensory loss; all others exhibited mixed sensory findings, involving both small and large fiber modalities. Most patients had extensive, overlapping involvement of multiple individual nerves, with legs more commonly affected than arms and distal nerves more commonly affected than proximal nerves (Table 1). The common peroneal was the most frequently involved nerve overall, and the ulnar nerve was most commonly affected in the arm. Despite the propensity for distally-accentuated damage (seen in all but two patients), proximal nerves were not routinely spared: 60% of patients had hip flexor weakness, 40% shoulder abductor and elbow flexor weakness, and 13% sensory loss in thoracic dermatomes. Only 8% of patients had cranial neuropathies, typically involving cranial nerve seven. All but one patient had asymmetric findings at presentation. Patterns of involvement were 77% asymmetric polyneuropathy (overlapping multifocal neuropathy), 13% true multiple mononeuropathy, 8% asymmetric lumbosacral plexopathy, and 2% distal symmetric polyneuropathy. Neurologic disability at presentation ranged from mild multifocal numbness and pain to severe quadriparesis and complete dependence. Median disability was 4 on the Prineas scale, consistent with partial dependence, a need for assistance with some bodily functions, and a preference for some type of walking aide. There was no significant difference in median disability or mean composite MRC score between the groups of patients with or without weight loss.

Laboratory findings

Laboratory results are detailed in Table E-1. Seventy-one percent of patients had an elevated ESR ( 20 mm/hour); in 25%, ESR was  50 mm/hour. Other laboratory markers of inflammation or autoantibody production were abnormal in a minority of patients, including positive ANA in 39%, anemia in 31%, leukocytosis in 23%, thrombocytosis in 21%, positive RF in 20%, and decreased complement (C3 or C4) in 11%. Nineteen patients underwent cerebrospinal fluid (CSF) examination; mild pleocytosis was noted in only one patient (5%). CSF protein was elevated (> 50 mg/dl) in 5/19 (26%), with a mean of 47.2 + 21.9 mg/dl and maximum of 106 mg/dl. Immunofixation electrophoresis revealed a MGUS in three patients (IgG, IgG, IgM), representing 8% (3/37) of patients older than 50 years.

Nerve biopsy was performed in all patients: sural nerve in 28, SPN/PBM in 18, and superficial radial nerve in two. Two patients with non-diagnostic SPN/PBM biopsies later had sural biopsies; conversely, one patient with a non-diagnostic sural biopsy subsequently underwent SPN/PBM biopsy. Thus, a total of 30 sural nerve, 19 SPN/PBM, and two superficial radial nerve biopsies were performed. Pathologic diagnoses and diagnostic sensitivities for these procedures are summarized in Table 2. SPN/PBM biopsy had a higher sensitivity for definite vasculitis than sural nerve biopsy, but the difference was not significant. In the 19 patients undergoing SPN/PBM biopsies, muscle revealed definite vasculitis in three patients (16%), inflammatory infiltrates in eight (42%), and muscle necrosis or regeneration in 13 (68%).


Treatment regimens and follow-up. Treatment and outcome measures are abstracted in Table E-2 on the Neurology Web site. In survivors of the initial 6 months of therapy, follow-up ranged from 6-260 months (median 63 months; mean 75 ± 54 months). Twenty-eight patients were initially treated with CS alone (prednisone, IV pulse methylprednisolone, or both) and 20 with combination therapy [CS plus an adjunctive agent – oral CYC in 18, azathioprine (AZA) in one, and intravenous immunoglobulin (IVIg) in one]. Starting prednisone doses ranged from 40-100 mg/day. The most common protocol consisted of prednisone 100 mg/day for 1-2 weeks, followed by 100 mg q.o.d. A starting dose of 100 mg q.o.d. was employed next most frequently. Starting CYC doses were 100 or 150 mg/day. In patients successfully tapered off prednisone (no progression or relapse), duration of therapy ranged from 2-45 months (median 12.5, mean 14.0 ± 8.8). In the 25 patients treated with CYC as a primary or rescue agent, duration of therapy ranged from 1-30 months (median 5 months). CYC had to be discontinued in 40% of patients because of adverse effects (pneumonia, cellulitis, sepsis, leukopenia, anemia, thrombocytopenia, rash, or diarrhea).

Comparison between initial CS monotherapy and combination therapy groups. For these two groups, baseline clinical and laboratory characteristics of potential relevance to therapeutic responses are charted in Table 3. There were no differences between the groups in any of these attributes. For survivors of the initial 6 months of therapy, median follow-up was 54.5 months (range 12-172, mean 61 ± 36 months) in the CS monotherapy group and 77 months (range 6-260, mean 94 ± 68 months) in the combination group. Methylprednisolone (1 gm/day IV for 3-6 days) preceded oral therapy in 57% of patients treated with CS alone and 65% of patients initially receiving combination therapy.

Outcome analyses are summarized in Table 4. For the primary outcome measure – long-term response rate – there was a significant difference favoring the initial combination therapy group. Of the secondary response variables, the combination therapy group also exhibited a significantly greater improvement in disability and a trend toward reduced incidence of chronic pain. Eighty-five percent (17/20) of patients initially treated with CS plus a second agent had improved disability at end of follow-up compared to 57% (16/28) of those receiving CS monotherapy. Long-term responders to CS monotherapy were twice as likely to relapse as responders to combination therapy, but this difference was not quite significant. The two groups had similar disease or initial treatment-related mortality rates and changes in composite MRC score. Kaplan-Meier survival plots for the CS monotherapy and combination therapy groups are depicted in the Figure. There was no significant difference between the groups. Estimated 5-year survival for the entire cohort was 87% (SE 13%).