Adult Primary Central Nervous System Vasculitis Treatment and Course

Transcription

1 ARTHRITIS & RHEUMATOLOGY Vol. 67, No. 6, June 2015, pp DOI /art VC 2015, American College of Rheumatology Adult Primary Central Nervous System Vasculitis Treatment and Course Analysis of One Hundred Sixty-Three Patients Carlo Salvarani, 1 Robert D. Brown Jr., 2 Teresa J. H. Christianson, 2 John Huston III, 2 Caterina Giannini, 2 Dylan V. Miller, 2 and Gene G. Hunder 2 1 Carlo Salvarani, MD: Mayo Clinic, Rochester, Minnesota, and Azienda Ospedaliera IRCCS di Reggio Emilia, Reggio Emilia, Italy; 2 Robert D. Brown Jr., MD, MPH, Teresa J. H. Christianson, BS, John Huston III, MD, Caterina Giannini, MD, Dylan V. Miller, MD, Gene G. Hunder, MD: Mayo Clinic, Rochester, Minnesota. Address correspondence to Carlo Salvarani, MD, Unita Operativa di Reumatologia, Azienda Ospedaliera IRCCS di Reggio Emilia, Viale Risorgimento N80, Reggio Emilia, Italy. salvarani.carlo@asmn.re.it. Submitted for publication August 14, 2014; accepted in revised form February 5, Objective. To describe the treatment and outcomes of patients with primary central nervous system (CNS) vasculitis. Methods. We retrospectively studied a cohort of 163 consecutive patients with primary CNS vasculitis who were seen at the Mayo Clinic over a 29-year period. We analyzed treatments, treatment responses, and factors predictive of outcomes. Results. A favorable response was observed in 85% of patients treated with prednisone alone and in 80% of patients treated with prednisone and cyclophosphamide. Relapses were observed in 27% of patients, and 25% of patients had discontinued therapy by the time of the last followup visit. Treatment with prednisone alone was associated with more frequent relapses (odds ratio [OR] 2.90), while large vessel involvement (OR 6.14) and cerebral infarcts at the time of diagnosis (OR 3.32) were associated with a poor response to treatment. Prominent gadolinium-enhanced cerebral lesions or meninges were linked with continued treatment at the last followup encounter (OR 2.28). Higher disability scores at the last followup visit were associated with increasing age at the time of diagnosis (OR 1.44) and cerebral infarctions (OR 3.74), while lower disability scores were associated with gadolinium-enhanced cerebral lesions or meninges (OR 0.35) and cerebral amyloid angiopathy (OR 0.24). Increased mortality was associated with increasing age at diagnosis (hazard ratio [HR] 1.39), diagnosis by angiography (HR 3.28), cerebral infarction (HR 4.44), and large vessel involvement (HR 4.98), while reduced mortality was associated with gadolinium-enhanced cerebral lesions or meninges (HR 0.20). Conclusion. The majority of patients with primary CNS vasculitis responded to treatment. Recognition of findings at diagnosis that predict the course or outcome may aid in decision-making regarding therapy. Primary central nervous system (CNS) vasculitis is an uncommon disorder of unknown cause that is restricted to the brain and spinal cord (1 3). Early case reports described primary CNS vasculitis as a fatal condition (1,4,5). In 1983, however, Cupps et al reported a favorable response to cyclophosphamide and glucocorticoids in 4 patients with primary CNS vasculitis (6). Subsequently, other investigators also reported a more favorable course with similar treatment (7 9). Because of the lack of uniform diagnostic criteria (2,10), however, and the relatively small numbers of subjects in most series, optimal management and treatment outcome remain uncertain. In addition, few studies have evaluated findings that may predict the response to treatment (7,8). In this study, we reviewed all cases of primary CNS vasculitis evaluated at the Mayo Clinic from 1983 to We analyzed the treatments and outcomes and assessed findings that may serve as guides for therapy. PATIENTS AND METHODS Identification of patients. In this study, we extended our earlier primary CNS vasculitis cohort of 101 consecutive patients seen at the Mayo Clinic over a 21-year period to 29 years, from 1983 to 2011 (7). The same predefined diagnostic 1637

2 1638 SALVARANI ET AL criteria were used (7) to include those examined at the Mayo Clinic from January 1, 2004 through December 31, During the more recent period, 62 additional patients were enrolled. Therefore, 163 patients with primary CNS vasculitis seen at the Mayo Clinic from 1983 to 2011 were included in this retrospective analysis. The study was approved by the Mayo Clinic Institutional Review Board. Patients were included if a brain or spinal cord biopsy sample showed vasculitis (transmural destructive inflammatory infiltrate) or if angiograms showed changes that were highly suggestive of vasculitis (smooth-wall segmental narrowing, dilatation, or occlusion affecting multiple cerebral arteries in the absence of proximal vessel changes consistent with atherosclerosis) (7). Angiograms were categorized into the following 2 groups: large/proximal artery (intracranial internal carotid and vertebral arteries, basilar artery, and proximal anterior, middle, and posterior cerebral arteries) and small/distal artery (intracranial artery and second division branches or smaller). Patients with vasculitis in organs other than the CNS and those with evidence of other diseases were excluded. None of the patients reported a history of exposure to vasoactive substances, were in the postpartum state, or had migraine headaches, thunderclap headaches, or manifestations typical of reversible cerebral vasoconstriction syndrome. Review of biopsy specimens and angiograms. Biopsy specimens were reviewed by 2 pathologists (CG and DVM), and angiograms were reviewed by a neuroradiologist. Conventional digital subtraction angiograms were the standard and were performed and interpreted according to a clinical protocol used by the Division of Neuroradiology at the Mayo Clinic. Clinical data collection. In cases with an uncertain initial diagnosis, the complete medical record was reviewed again by 2 rheumatologists (CS and GGH) and 1 neurologist (RDB) to reach a consensus. Comprehensive information was recorded about all medical conditions (previous, at the time of admission, and at followup), all tests and procedures, and all treatments and outcomes. All patients underwent a complete neurologic examination performed by a neurologist at the time of diagnosis and on subsequent visits. Definition of relapse. Relapse was defined as a recurrence or worsening of symptoms of primary CNS vasculitis or progression of existing or evidence of new lesions on subsequent magnetic resonance imaging (MRI) examinations while the patient was receiving no medication or a stable dosage of medication. Patients who experienced a relapse required additional therapy. To assess treatment, we used the treating physician s global opinion about response, based on a detailed review of the medical record. The degree of disability at presentation and at the last visit was defined by a review of the medical record and was categorized using the modified Rankin scale. The scale consists of 7grades(0to6),wheregrade05 no neurologic signs or symptoms, grade 1 5 no significant disability, grades increasing disability, and grade 6 5 death (7,11). The patients were followed up until death or the last followup visit. Statistical analysis. Numeric parameters were compared using a 2-sided 2-sample t-test or a Wilcoxon s rank sum test when the distributions were skewed. Categorical variables were compared by chi-square test or Fisher s exact test when cell counts were small. Logistic regression models were used to identify characteristics at diagnosis that increased the odds of a poor outcome, relapses, treatment response, and inability to discontinue treatment at the last followup. Univariate and age-adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) were reported. A Cox proportional hazards model was used to assess the relationship between demographic, clinical, laboratory, radiologic, pathologic, and therapeutic parameters at diagnosis and survival. Univariate and age-adjusted hazard ratios (HRs) and 95% CIs were reported. Results were reported as ageadjusted when age was significantly associated with the outcome. Multivariable modeling, adjusted for age, was performed manually in a forward selection manner because the risk factors under consideration had differing patterns of missing data for the patients. Because of the limited number of events, the multivariable model could support no more than 2 additional risk factors. All P values were 2-sided. P values less than 0.05 were considered significant. Statistical analysis was performed using SAS software, version 9. In some instances, Rankin scores were grouped variously to summarize the results. Rankin scores were dichotomized into 0-3 and 4-5, because it was medically relevant. Patients with a Rankin score of 0-3 maintain some degree of independence in the activities of daily living, while patients with ascoreof.3 need complete assistance. A Rankin score of 6 represents death. RESULTS Table 1 shows the characteristics of the 163 patients at diagnosis. The median duration of followup was 12 months (range years). Brain or spinal cord tissue specimens were obtained from 81 patients. Vasculitis was found on biopsy in 58 patients (72%); the biopsy results in the remaining 23 patients were negative. A granulomatous inflammatory histologic pattern was observed in 34 patients (59%) (accompanied by vascular deposits of b-amyloid peptide in 20 [34%]), a granulomatous and necrotizing pattern in 1 patient (2%), an acute necrotizing pattern in 10 patients (17%), and a lymphocytic pattern in 13 patients (22%). Angiography was performed in 129 patients, and the results were consistent with vasculitis in 113 (88%). Angiograms alone were used to confirm the diagnosis of vasculitis in 105 patients, including the 23 with negative biopsy results, and also showed vasculitis in 8 patients in whom the results of brain biopsy were positive. No differences in clinical manifestations at the time of diagnosis, laboratory investigations, including spinal fluid analysis, and histopathologic patterns were observed in the new ( ) cohort and the earlier ( ) cohort (Table 1). The new cohort had a higher median age at diagnosis (P ), a higher frequency of gadoliniumenhanced meninges on MRI (P ), a reduced frequency of patients with large vessel changes on angiography (P ), and a higher percentage of patients in whom magnetic resonance angiography was performed within 3 months of diagnosis (61.3% versus 31.7%; P ). Also, the new cohort had a higher level of disability (Rankin score of 4/5) at presentation (43.5% versus 21.8%; P ) and at the last followup visit (14.5%

3 THERAPY AND OUTCOMES IN PRIMARY CNS VASCULITIS 1639 Table 1. Findings at the time of diagnosis in the patients with primary central nervous system vasculitis, according to the study period* Characteristic All patients (n 5 163) cohort (n 5 101) cohort (n 5 62) Male sex 72 (44.2) 43 (42.6) 29 (46.8) Age at diagnosis, median (range) years 48 (17 85) 47 (17 84) 51.5 (20 85) Years from symptom onset to diagnosis, median (range) 0.1 ( ) 0.1 ( ) 0.1 ( ) Clinical manifestations at presentation Headache 97 (59.5) 64 (63.4) 33 (53.2) Cognitive dysfunction 88 (54) 50 (49.5) 38 (61.3) Persistent neurologic deficit or stroke 66 (40.5) 40 (39.6) 26 (41.9) Seizures 33 (20.2) 16 (15.8) 17 (27.4) Intracranial hemorrhage 16 (9.8) 8 (7.9) 8 (12.9) Systemic manifestations 15 (9.2) 9 (8.9) 6 (9.7) Fever 16 (9.8) 9 (8.9) 7 (11.3) CSF abnormality Protein.70 mg/dl (normal 14 15) 63 (52.1) 38 (52.1) 25 (52.1) Protein.45 mg/dl or WBC count.5/mm (81.3) 56 (75.7) 44 (89.8) Protein.70 mg/dl or WBC count.10/mm 3 77 (63.6) 46 (63) 31 (64.6) ESR, median (range) mm/hour 8 (0 124) 9 (0 110) 7.5 (0 124) Initial MRI findings Infarcts 81 (54.4) 48 (53.3) 33 (55.9) Gd-enhanced lesions (intracranial or meningeal) 60 (40.3) 33 (36.7) 27 (45.8) Gd-enhanced meningeal lesions 29 (19.5) 11 (12.2) 18 (30.5) Angiographic findings# Large/proximal vessel vasculitis 75 (65.8) 55 (72.4) 20 (52.6) Small/distal vessel vasculitis 103 (90.4) 70 (92.1) 33 (86.8) * Except where indicated otherwise, values are the number (%) of patients. ESR 5 erythrocyte sedimentation rate; Gd 5 gadolinium. P, 0.05 versus cohort. Defined as the presence of at least 1 of the following: fatigue, anorexia, weight loss, arthralgia. Cerebrospinal fluid (CSF) data were available for 121 patients for protein.70 mg/dl, 123 patients for protein.45 mg/dl or white blood cell (WBC) count.5/mm 3, and 121 patients for protein.70 mg/dl or WBC count.10/mm 3. Magnetic resonance imaging (MRI) was performed in 149 patients. # Cerebral angiography was performed in 129 patients. Information on vasculitis type (large vessel and small vessel) was available for 113 patients. versus 3%; P ), but mortality at the last followup was not different between the 2 cohorts (12.9% in the new cohort versus 16.8% in the earlier cohort) nor was the frequency of a bad outcome (Rankin score of 4 6) at the last followup (27.4% versus 19.8%; P ). Initial treatment. The details of initial treatment are shown in Table 2. One hundred fifty-nine patients received treatment, and 4 patients did not receive treatment. Among the treated patients, 157 received glucocorticoids. In 66 patients, intravenous (IV) pulse methylprednisolone therapy (3 17 pulses [median 5 pulses], usually 1 gm/pulse) preceded oral prednisone therapy. The median starting dosage of oral prednisone in the 157 treated patients was 60 mg/day. The median duration of oral prednisone therapy was 9 months (range months). Three-fourths of the patients were treated for #17 months. In 75 patients, prednisone was the only therapeutic agent used as initial treatment. In 82 patients, prednisone was combined with a second drug as initial therapy; 72 of these patients received prednisone plus cyclophosphamide (2 patients were treated with cyclophosphamide alone). Fifty-one patients initially received treatment orally. The median initial dosage of oral cyclophosphamide was 150 mg/day (range mg/day), and the median duration of therapy was 7 months (range 1 33 months). Three-fourths of the patients were treated for #12 months. Twenty-three patients received intermittent IV pulse cyclophosphamide. The median dosage was 1,000 mg/month, and the median length of the treatment was 2 months (range 1 17 months). Three-fourths of the patients received #6 monthly pulses. Among the 75 patients initially treated with prednisone alone, the median duration of all therapy was 11.1 months (range months). Among the 72 patients initially treated with prednisone plus cyclophosphamide, the median duration of all therapy was 11.6 months (range months). Six of the 82 patients were initially treated with prednisone plus azathioprine at a median initial dosage of 100 mg/day (range mg/day). The median duration of treatment was 11 months (range months). Three patients initially received prednisone plus mycophenolate mofetil at a median dosage of 2,000 mg/day, and 1 patient had received 2 IV injections of rituximab. Aspirin was started at the time of diagnosis in 24 (15%) of the 163 patients. The role of

4 1640 SALVARANI ET AL Table 2. Treatment in the patients with primary central nervous system vasculitis, according to study period* All patients (n 5 163) cohort (n 5 101) cohort (n 5 62) Initial treatment Prednisone alone 75 (46) 48 (47.5) 27 (43.5) Prednisone and cyclophosphamide (oral/pulse) 72 (44) 46 (45.5) 26 (41.9) Prednisone and azathioprine 6 (3.7) 4 (4.0) 2 (3.2) Prednisone and other therapy 4 (2.5) 0 4 (6.5) Cyclophosphamide alone 2 (1.2) 2 (2.0) 0 No therapy 4 (2.5) 1 (1.0) 3 (4.8) Treatment duration, median (range) months Prednisone 9.0 ( ) 9.5 ( ) 6.0 ( ) Cyclophosphamide (oral/pulse) 7.0 ( ) 10.0 ( ) 6.0 ( ) Patients with relapses/total cohort 44/163 (27.0) 26/101 (25.7) 18/62 (29.0) Prednisone alone 29/75 (38.7) Prednisone and cyclophosphamide (oral/pulse) 13/72 (18.1) Patients not requiring therapy at last followup 40/158 (25.3) 28/99 (28.3) 12/59 (20.3) Prednisone alone 21/75 (28.0) Prednisone and cyclophosphamide (oral/pulse) 18/72 (25.0) Patients responding to the treatment 126/152 (82.9) 80/97 (82.5) 46/55 (83.6) Prednisone alone 62/73 (84.9) Prednisone and cyclophosphamide (oral/pulse) 55/69 (79.7) * No significant differences between the 2 cohorts were observed. Except where indicated otherwise, values are number of patients/number of patients assessed (%). Including 3 patients treated with mycophenolate mofetil and 1 treated with rituximab. Defined by the clinical judgment of the treating physician. aspirin in treatment could not be evaluated, but it was not significantly associated with a poor outcome, survival, or intracranial hemorrhage. Response to initial therapy. Adequate information to judge the response to initial therapy was available in the medical records of 152 patients. Overall, a favorable response was observed in 126 (83%) of these patients. Sixty-two (85%) of the 73 patients followed up who were receiving prednisone alone as initial therapy responded favorably, and 55 (89%) of these patients had experienced improvement within the first 2 months. Fifty-five (80%) of 69 patients followed up who were receiving prednisone plus cyclophosphamide responded favorably, and 46 (84%) of these patients had improved within 2 months. The differences between the patients treated with prednisone alone and those treated with prednisone plus cyclophosphamide were not significant. Followup information was available for 5 of the 6 patients who received initial treatment with prednisone plus azathioprine. Four of these patients experienced improvement with therapy, and all did so within 1 month. A favorable response to therapy was observed in 80% of patients with lymphocytic vasculitis, 100% with necrotizing vasculitis, and 100% with granulomatous vasculitis (P ). Followup treatment. In 5 patients initially treated with prednisone alone, mycophenolate mofetil therapy was started within 3 months of beginning therapy. For the analyses, these 5 patients were added to the 3 who initially received prednisone plus mycophenolate mofetil. Followup information was available for 7 of these 8 patients. All 7 patients had a favorable response within 4 months, and 6 had a favorable response within 2 months. Overall, 91 patients were given cyclophosphamide at some time during therapy; 74 received cyclophosphamide as initial therapy, and 17 received cyclophosphamide to treat relapses. Among the 55 patients who were initially treated with prednisone plus cyclophosphamide and showed improvement at followup, 11 were changed from cyclophosphamide to azathioprine, 3 to mycophenolate mofetil, and 3 to methotrexate (15 20 mg/week) in an attempt to reduce the risk of toxic effects of cyclophosphamide. All 17 of these patients who discontinued cyclophosphamide had Rankin scores in the 0 3 range at last followup, whereas 31 (82%) of 38 patients who were still receiving cyclophosphamide at the last followup visit had Rankin scores in the 0 3 range (P ). Two patients were treated with plasma exchange, and 1 patient was treated with infliximab for disease refractory to glucocorticoids plus cyclophosphamide. Fifty-three patients received only glucocorticoids throughout the course of therapy. The median duration of treatment in these patients was 6.0 months (range months). Twenty-two other patients were given only prednisone initially, but an immunosuppressant was added later. The median duration of therapy in these patients was 22 months, (range months). Eighty-two

5 THERAPY AND OUTCOMES IN PRIMARY CNS VASCULITIS 1641 patients received prednisone plus an immunosuppressant drug at diagnosis. The median duration of treatment in these patients was 11.8 months (range months). Occurrence and treatment of relapses. Relapses occurred in 44 (28%) of the 159 treated patients. Twentyeight of these patients had 1 relapse, 10 had 2 relapses, and 6 had $3 relapses. Patients who experienced a relapse had a longer median duration of treatment compared with those who had no relapse (18 months versus 9 months; P, 0.001). The medical records showed no evidence that relapses were related to rapid withdrawal of therapy. Treatment of relapses included cyclophosphamide in 24 patients (oral treatment in 15 and IV treatment in 9), methylprednisolone pulses in 6 patients, mycophenolate mofetil in 3 patients, chlorambucil in 3 patients, and plasma exchange in 1 patient. One patient was treated with etanercept for a relapsing course despite previous treatment with cyclophosphamide and mycophenolate mofetil. The remaining patients received increased doses of oral glucocorticoids. All patients appeared to improve or stabilize in response to the treatments for relapse. Relapses occurred more frequently in patients treated initially with prednisone alone compared with those treated with prednisone plus cyclophosphamide (39% versus 18%; P ) (Table 2). Relapses were observed in 22 (31%) of 71 patients with large vessel or large vessel/small vessel involvement and in 12 (24%) of 51 patients with only small vessel changes or angiography-negative biopsypositive disease; the difference was not statistically significant (P ). Clinical factors influencing response to treatment. Forty (25%) of 158 patients had discontinued therapy by the time of the last followup visit; no differences were observed between patients treated initially with prednisone alone and those treated initially with prednisone plus cyclophosphamide. Twenty-one (28%) of the 75 patients receiving prednisone alone and 18 (25%) of 72 of those treated with prednisone plus cyclophosphamide were not receiving therapy at the last followup visit. Univariate logistic modeling was used to assess an association of findings at diagnosis with response to treatment, relapses, and therapy at last followup. Large vessel involvement (OR 6.14, 95% CI , P ) and cerebral infarcts on MRI at diagnosis (OR 3.32, 95% CI , P ) were associated with a poor response to treatment, while prominent gadolinium-enhanced cerebral lesions or meninges assessed by MRI were associated with longer therapy (OR 2.28, 95% CI , P ), which was often being continued at the time of last followup. Initial treatment with prednisone alone was the only finding significantly associated with relapses (OR 2.90, 95% CI , P ). Additional multivariate analysis was not possible because of small numbers. No significant differences were observed between the group with biopsy-diagnosed disease and the group with angiographically diagnosed disease, including median duration of therapy (11.7 months versus 11.2 months), number of patients with relapses (18 [33%] of 54 versus 26 [26%] of 99), response to treatment (46 [87%] of 53 versus 80 [81%] of 99), and patients not requiring therapy at last followup (11 [20%] of 55 versus 29 [28%] of 103). The patients with angiographically diagnosed disease were significantly more frequently treated with cyclophosphamide than with prednisone alone (55 [52%] of 105 versus 19 [33%] of 58; P ). Cerebral amyloid-b related angiopathy (ABRA). We compared the 20 patients whose CNS biopsies showed vasculitis with amyloid deposits (ABRA) with the 38 patients whose biopsies showed vasculitis without amyloid deposits. No significant differences were observed between the 2 groups regarding the numbers treated with IV pulse glucocorticoids (11 [55%] of 20 versus 14 [37%] of 38), oral prednisone (19 [95%] of 20 versus 36 [95%] of 38), daily oral cyclophosphamide (5 [25%] of 20 versus 7 [18%] of 38), and IV pulse cyclophosphamide (3 [15%] of 20 versus 4 [11%] of 38). The median duration of therapy was shorter in patients with ABRA (6 months versus 15 months; P ). No significant differences were observed between patients with ABRA and those without amyloid deposits regarding the frequency of those who experienced relapses (5 [26%] of 19 versus 13 [37%] of 35), who responded to treatment (16 [84%] of 19 versus 30 [88%] of 34), and who had discontinued therapy at last followup (3 [16%] of 19 versus 8 [22%] of 36). Status at last followup. Table 3 compares the Rankin disability scores at presentation with those at last followup. As shown, the scores at diagnosis were divided into 3 groups of increasing disability (0 2, 3, and 4 5). The scores at last followup were also divided into 3 groups (0 3, 4 5, and 6 [death]). The duration of followup was separated into 3 intervals. Of the 114 patients with low/intermediate disability scores at diagnosis (score of 0 3), 101 had low/ intermediate scores at the last followup visit. Half of the 49 patients with severe disability at diagnosis (score of 4 5) had less disability at followup (score of 0 3). Twenty-five (15.3%) of 163 patients died during followup. Most deaths occurred within the first year, especially those in patients presenting with severe disease (score of 4 5). However, some deaths occurred during all followup intervals. High disability scores (Rankin score of 4 6) were equally frequent at last followup in patients with relapsing disease (10 [23%] of 44) and those without relapses (27 [23%] of 119). No significant differences in

6 1642 SALVARANI ET AL Table 3. Rankin scores at diagnosis and last followup Score at diagnosis Score at followup 0 2 (n 5 83) 3 (n 5 31) 4 5 (n 5 49) Followup duration,1 year (deceased) Followup duration years (deceased) Followup duration 5 15 years (deceased) * Values are the number of patients. the number of deaths during followup were observed between patients with relapse and those without relapse (8 [18%] of 44 versus 17 [14%] of 119). High disability scores at last followup were more frequent in patients with angiography-diagnosed disease than in patients in whom the diagnosis was made by biopsy (27 [26%] of 105 versus 10 [17%] of 58), but the difference was not significant. Univariate logistic analysis adjusted for age was also used to assess the association of specific findings at diagnosis with the Rankin score at the last followup (data not shown). High disability scores (Rankin score of 4 6) at last followup were significantly associated with increasing age (calculated per 10-year increment) (OR 1.44, 95% CI , P ) and cerebral infarction at diagnosis assessed by MRI (OR 3.74, 95% CI , P ). Patients with prominent gadolinium-enhanced cerebral lesions or meninges (OR 0.35, 95% CI , P ) and patients with ABRA (OR 0.24, 95% CI , P ) had lower disability scores at the last followup. Other factors were not associated with a high Rankin score, including treatment with prednisone alone or cyclophosphamide plus prednisone. In a multivariable model, increasing age (per 10- year increment) (OR 1.47, 95% CI , P ) and MRI findings of infarcts (OR 3.41, 95% CI , P ) were associated with high disability scores (Rankin score of 4 6) at the last followup, while MRI findings of gadolinium-enhanced cerebral lesions or meninges (OR 0.40, 95% CI , P ) were associated with less disability. A univariate Cox proportional hazards model was used to assess the association between increased mortality and findings at diagnosis (Table 4). Increasing age (calculated per 10-year increments) (HR 1.39), method of diagnosis (angiography versus biopsy) (HR 3.28), cerebral infarction versus no infarction (HR 4.44), and large vessel involvement versus small vessel involvement (HR 4.98) were associated with increased mortality. Patients with prominent gadolinium-enhanced cerebral lesions or meninges (by MRI) had a lower risk of death (HR 0.20) than patients with no such lesions at presentation. Patients with ABRA also had a lower risk of death (HR 0.17), although the difference was not statistically significant. No differences in mortality were observed when patients were Table 4. Characteristics associated with increased mortality* Characteristic HR (95% CI) Univariate P Multivariate HR (95% CI) Age, per 10-year difference 1.39 ( ) ( ) Male vs. female 0.80 ( ) 0.61 Main symptom at presentation Headache or constitutional symptom 1.00 Focal manifestation vs. headache or constitutional symptom 2.42 ( ) 0.17 Cognitive disorder vs. headache or constitutional symptom 3.40 ( ) Diagnosis by angiography only vs. biopsy 3.28 ( ) MRI findings Infarct vs. no infarct 4.44 ( ) ( ) Gd-enhanced lesions or meninges vs. normal or minimal changes 0.20 ( ) ( ) Large vessel involvement vs. small vessel involvement 4.98 ( ) 0.01 Increased CSF protein level (.70 mg/dl) 1.29 ( ) 0.61 Cerebral amyloid angiopathy, present vs. absent 0.17 ( ) Prednisone alone vs. cyclophosphamide and prednisone 1.03 ( ) 0.94 Rapid (,1 month) vs. slow (.1 month) onset 1.27 ( ) 0.57 * Univariate and multivariate Cox proportional hazards models were used for age-adjusted analysis. HR 5 hazard ratio; 95% CI 5 95% confidence interval; MRI 5 magnetic resonance imaging; Gd 5 gadolinium; CSF 5 cerebrospinal fluid. Data were available for 129 patients.

7 THERAPY AND OUTCOMES IN PRIMARY CNS VASCULITIS 1643 stratified according to treatment (prednisone alone versus prednisone plus cyclophosphamide). In a multivariate model (Table 4), increasing age (per 10-year increment) (P ) and MRI findings of infarcts (P ) were associated with increased mortality, and MRI findings of gadolinium-enhanced cerebral lesions or meninges (P ) were associated with lower mortality. Among patients with biopsy-proven disease, the outcomes varied according to the histologic pattern, but the differences were not significant. All 13 patients with a lymphocytic pattern had a good outcome at the last followup (Rankin score of 0 3). Eighty percent of the 35 patients with a granulomatous pattern and 70% of the 10 patients with a necrotizing pattern also had a good outcome. We observed a trend toward a higher frequency of patients with a good outcome at last followup in the combined group with lymphocytic vasculitis compared with the groups with granulomatous and necrotizing vasculitis (100% and 77.8%, respectively; P ). None of the patients in the group with lymphocytic vasculitis had died by the time of last followup compared with 11.1% of patients in the other 2 groups (P ). DISCUSSION The majority of patients in the study responded to either prednisone alone or to prednisone plus cyclophosphamide in doses used to treat other vasculitides (12). The response rate for both regimens was similar, and the majority of patients showed improvement within 2 months. The clinical findings at diagnosis were comparable in the 2 treatment groups, as were the Rankin disability scores at diagnosis and last followup. The response to treatment was also similar in patients in whom vasculitis was diagnosed by biopsy and those in whom the diagnosis was made by angiography. The only difference between therapy results was a higher relapse rate in the group treated with prednisone alone, and in that sense, the 2 regimens were not completely equivalent. However, more than half of the patients (72%) achieved a sustained therapeutic response (no relapses) during followup. Patients with relapses needed longer therapy compared with those without relapses, but relapse was not associated with increased mortality or higher disability (Rankin score) at the last followup visit. Mortality during the followup period was 15% (25 of 163 patients), which indicates the seriousness of this disease in spite of a favorable response in many. Mortality in our cohort was higher than that observed in a recent French series of 52 patients (6%) (8). In the French series,.80% of patients received glucocorticoids combined with cyclophosphamide, while in our series this combination was prescribed as initial therapy in only 44% of patients. As noted above, however, patients treated with prednisone alone in our series did not have increased mortality or a higher Rankin score at last followup compared with those treated with combination therapy. Therefore, the higher mortality observed in our study was not likely related to less aggressive treatment. Differences in the population of patients with primary CNS vasculitis enrolled in the 2 studies may explain the disparity. In our study, the majority of patients with biopsyproven primary CNS vasculitis had a granulomatous or necrotizing pattern, while lymphocytic vasculitis was the prevalent pattern in the French cohort. In our cohort, we notedanassociationbetweenincreasedmortalityandlarge vessel lesions and CNS infarctions on MRI, regardless of the type of therapy. In such cases, the CNS biopsy showed a predominant granulomatous and/or necrotizing pattern (13,14). Furthermore, we showed that lymphocytic vasculitis was associated with a better outcome and lower mortality compared with the other 2 patterns. Therefore, differences in disease severity linked to differences in the histopathologic patterns may explain the higher mortality in our series. The most severe end of the clinical spectrum of primary CNS vasculitis is represented by patients with angiographic evidence of multiple bilateral large vessel lesions and MRI evidence of multiple cerebral infarctions. These patients had a rapidly progressive clinical course and responded poorly to therapy (13,14). In such patients, it is reasonable to start more aggressive treatment with prednisone and cyclophosphamide. Patients with small vessel involvement appear to have less-severe disease. These patients responded to therapy and had better survival and less disability at followup. Among these patients are those with angiographynegative, biopsy-positive primary CNS vasculitis. In these patients, vessel lesions are beyond the resolution of angiography, and a biopsy is required for diagnosis (15 18). Such patients often presented with cognitive dysfunction and had highly elevated concentrations of CSF proteins and gadolinium-enhanced cerebral lesions or meninges onmri.patientswithabracouldbeincludedinthis subset (15,16). Glucocorticoids might be the appropriate treatment in these patients, with cyclophosphamide or a different immunosuppressant agent reserved for the treatment of relapses. The relatively large number of patients who received glucocorticoids alone throughout the course of their illness supports this idea. We also observed that patients with gadoliniumenhanced cerebral lesions or meninges required longer therapy and were still receiving treatment at the last followup visit. These MRI-visualized lesions were also

8 1644 SALVARANI ET AL observed in patients at higher risk of relapses (OR 1.49), although the association was not significant. Our data partially confirm the recent observation by de Boysson et al that gadolinium-enhanced meninges on MRI were associated with higher relapse rates (8), but we did not observe an association between clinical manifestations at diagnosis and relapses. Almost one-fifth of the patients treated with both glucocorticoids and immunosuppressive agents as initial treatment were treated with a second drug different from cyclophosphamide (mainly azathioprine or mycophenolate mofetil). Most of these patients showed a favorable response, suggesting that these drugs can replace the more toxic cyclophosphamide for induction of remission. Seventeen (24%) of the 72 patients treated initially with cyclophosphamide were switched to a different immunosuppressive agent for maintenance therapy. Azathioprine was used most often, followed by mycophenolate mofetil and methotrexate. These drugs appeared to be effective, because improvement was maintained. In other words, cyclophosphamide did not appear essential to the improved outcome of the disease. It could also be noted that although aspirin therapy had no noted beneficial effect, it was not associated with a harmful effect such as increased subsequent intracranial hemorrhage. Because of the low number of patients treated, however, further studies are needed to define the real benefit of these drugs. Three patients were treated with biologic agents, and all had a favorable response. One patient was given rituximab because she declined treatment with cyclophosphamide. Two other patients were treated with tumor necrosis factor a blockers (for disease refractory to treatment and for relapsing vasculitis, respectively). Therefore, although the experience is limited to few case reports (19 21), biologic agents may represent a useful therapeutic option for patients who are intolerant to conventional immunosuppressive agents or as second-line therapy for patients with treatment-refractory disease or relapses. The current study had several limitations. As in all retrospective studies, incomplete data sets may have influenced the findings. Other limitations include the possibility of referral bias of cases, a relatively short followup period, and the lack of tissue at diagnosis in the majority of patients, even though followup of this group tended to confirm the diagnosis of primary CNS vasculitis. A lack of uniformity of treatment regimens was also a limitation. Strengths of the study were the large number of unselected consecutive cases defined by uniform pathologic criteria, the extensive clinical data available, and followup information. AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Salvarani had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study conception and design. Salvarani, Brown, Christianson, Huston, Giannini, Miller, Hunder. Acquisition of data. Salvarani, Brown, Christianson, Huston, Giannini, Miller, Hunder. Analysis and interpretation of data. Salvarani, Brown, Christianson, Huston, Giannini, Miller, Hunder. REFERENCES 1. Salvarani C, Brown RD Jr, Hunder GG. Adult primary central nervous system vasculitis. Lancet 2012;380: Calabrese LH, Mallek JA. Primary angiitis of the central nervous system: report of 8 new cases, review of the literature, and proposal for diagnostic criteria. Medicine (Baltimore) 1988;67: Salvarani C, Brown RD Jr, Hunder GG. Adult primary central nervous system vasculitis: an update. Curr Opin Rheumatol 2012;24: Cravioto H, Feigin I. Noninfectious granulomatous angiitis with a predilection for the nervous system. Neurology 1959;9: Lie JT. Primary (granulomatous) angiitis of the central nervous system: a clinicopathologic analysis of 15 new cases and a review of the literature. Hum Pathol 1992;23: Cupps TR, Moore PM, Fauci AS. Isolated angiitis of the central nervous system: prospective diagnostic and therapeutic experience. Am J Med 1983;74: Salvarani C, Brown RD Jr, Calamia KT, Christianson TJ, Weigand SD, Miller DV, et al. Primary central nervous system vasculitis: analysis of 101 patients. Ann Neurol 2007;62: De Boysson H, Zuber M, Naggara O, Neau JP, Gray F, Bousser MG, et al, for the French Vasculitis Study Group and the French NeuroVascular Society. Primary angiitis of the central nervous system: description of the first fifty-two adults enrolled in the French cohort of patients with primary vasculitis of the central nervous system. Arthritis Rheumatol 2014;66: Oon S, Roberts C, Gorelik A, Wicks I, Brand C. Primary angiitis of the central nervous system: experience of a Victorian tertiaryreferral hospital. Intern Med J 2013;43: Birnbaum J, Hellmann DB. Primary angiitis of the central nervous system. Arch Neurol 2009;66: Sulter G, Steen C, De Keyser J. Use of the Barthel index and modified Rankin scale in acute stroke trials. Stroke 1999;30: Molloy ES, Langford CA. Advances in the treatment of small vessel vasculitis. Rheum Dis Clin North Am 2006;32: Salvarani C, Brown RD Jr, Calamia KT, Christianson T, Huston J III, Meschia JF, et al. Rapidly progressive primary central nervous system vasculitis. Rheumatology (Oxford) 2008;50: Salvarani C, Brown RD Jr, Morris JM, Huston J III, Hunder GG. Catastrophic primary central nervous system vasculitis. Clin Exp Rheumatol 2014;32:S Salvarani C, Brown RD Jr, Calamia KT, Christianson TJ, Huston J III, Meschia JF, et al. Primary central nervous system vasculitis: comparison of patients with and without cerebral amyloid angiopathy. Rheumatology (Oxford) 2008;47: Salvarani C, Hunder GG, Morris JM, Brown RD Jr, Christianson T, Giannini C. Ab-related angiitis: comparison with CAA without inflammation and primary CNS vasculitis. Neurology 2013;81:

9 THERAPY AND OUTCOMES IN PRIMARY CNS VASCULITIS Salvarani C, Brown RD Jr, Calamia KT, Christianson TJ, Huston J III, Meschia JF, et al. Angiography-negative primary central nervous system vasculitis: a syndrome involving small cerebral vessels. Medicine (Baltimore) 2008;87: Salvarani C, Brown RD Jr, Calamia KT, Christianson TJ, Huston J III, Meschia JF, et al. Primary central nervous system vasculitis with prominent leptomeningeal enhancement: a subset with a benign outcome. Arthritis Rheum 2008;58: Salvarani C, Brown RD Jr, Calamia KT, Huston J III, Meschia JF, Giannini C, et al. Efficacy of tumor necrosis factor a blockade in primary central nervous system vasculitis resistant to immunosuppressive treatment. Arthritis Rheum 2008;59: Salvarani C, Brown RD Jr, Huston J III, Morris JM, Hunder GG. Treatment of primary CNS vasculitis with rituximab: case report. Neurology 2014;82: De Boysson H, Arquizan C, Guillevin L, Pagnoux C. Rituximab for primary angiitis of the central nervous system: report of 2 patients from the French COVAC cohort and review of the literature. J Rheumatol 2013;40: DOI /art Erratum In the editorial by Wolfe published in the February 2015 issue of Arthritis & Rheumatology (pages ), Dr. Wolfe made several statements that have been challenged. In particular, he stated that a study published by Bennett et al (Arthritis Care Res [Hoboken] 2014;66: ) failed to evaluate controls for the presence of fibromyalgia, and thus the published sensitivity and specificity of these criteria were incorrect. This was not the case as Bennett et al did evaluate controls for fibromyalgia in their study. The editorial also referred to new criteria for fibromyalgia that Dr. Wolfe and colleagues published in 2010 as the 2010 American College of Rheumatology (ACR) fibromyalgia criteria. However, the accurate designation for these criteria is the 2010 American College of Rheumatology preliminary diagnostic criteria for fibromyalgia. We regret the errors.